diff --git a/bundles/showreel/build_showreel.py b/bundles/showreel/build_showreel.py index 6ed5551..85597da 100644 --- a/bundles/showreel/build_showreel.py +++ b/bundles/showreel/build_showreel.py @@ -29,7 +29,7 @@ _ROOT = os.path.dirname(os.path.dirname(_HERE)) exec(compile(open(os.path.join(_ROOT, 'lib', 'tox_build.py')).read(), 'tox_build.py', 'exec'), globals()) -VERSION = '0.1.29' +VERSION = '0.1.30' # Shown while nothing is wired: the dancer, the library's own footage (CC BY 4.0, 5 s, 1.3 MB), and # her depth map (Depth Anything 3 Mono-Large, held steady across frames, 0.5 MB) for the effects that # take input 3, carried in the component's Virtual File System and read by a path relative to it, so diff --git a/bundles/showreel/review.md b/bundles/showreel/review.md index 5c96397..8fe1804 100644 --- a/bundles/showreel/review.md +++ b/bundles/showreel/review.md @@ -1,35 +1,35 @@ # showreel: idiom review -Reviewed version: 0.1.29 +Reviewed version: 0.1.30 The review items of docs/TD_IDIOMS.md, checked by hand for this version: tick each box (`- [x]`) and say in a few words how it holds. - [x] P4 Pulse for actions, Toggle for states, Menu for discrete choices; labels in Title Case. Names of 10 characters or fewer are checked at save time (assert_idiomatic); TouchDesigner itself refuses names that aren't an uppercase letter followed by lowercase letters or digits. [Custom Parameters](https://docs.derivative.ca/Custom_Parameters), [Page Class](https://docs.derivative.ca/Page_Class) - Menu for View (discrete choices) and Flashing; Toggle for Labels and Full Screen; Int for Every and Display; Pulses for Open Window and Open Docs. Labels in Title Case; names are 10 characters or fewer (save-time check passes). The effects inside keep their own reviewed pages. 0.1.3 adds Fluid inside (its own page of parameters, reviewed in effects/fluid/review.md); the showreel's page is unchanged. 0.1.4 adds VHS + CRT inside (its own page, reviewed in effects/vhs_crt/review.md); the showreel's page is unchanged. 0.1.5: VHS + CRT 0.2.0 inside (its page gains Screen and the screens' settings, reviewed in effects/vhs_crt/review.md); the showreel's page is unchanged. 0.1.6: the effects inside are rebuilt for the clock fix (Blob Track 0.1.3, Datamosh 0.2.1, Fluid 0.1.1, Particles 0.3.2, Point Cloud 0.1.3, VHS + CRT 0.2.1); the showreel's page is unchanged. 0.1.7 adds Bent Cam inside (its own page, reviewed in effects/bent_cam/review.md), and every effect inside is rebuilt for the shared red guard's move into lib/glsl; the showreel's page is unchanged. 0.1.8: Bent Cam 0.1.1 inside (its page gains Beat Kick, reviewed in effects/bent_cam/review.md); the showreel's page is unchanged. 0.1.9: Slit-Scan inside (its own page, reviewed in effects/slit_scan/review.md); the showreel's page is unchanged. 0.1.10: Pixel Sort inside (its own page, reviewed in effects/pixel_sort/review.md); the showreel's page is unchanged. 0.1.11: Paint inside (its own page, reviewed in effects/paint/review.md), and every effect inside rebuilt for lib's new structure tensor; the showreel's page is unchanged. 0.1.13: Reaction-Diffusion inside (its own page, reviewed in effects/reaction_diffusion/review.md), and every effect inside rebuilt for lib's structure_tensor(ease=); the showreel's page is unchanged. 0.1.14: Pencil inside (its own page, reviewed in effects/pencil/review.md), and every effect inside rebuilt for lib's paper.glsl (moved from Paint unchanged); the showreel's page is unchanged. 0.1.15: Motion Magnify inside (its own page, reviewed in effects/motion_magnify/review.md); the showreel's page is unchanged. 0.1.16: Video Vocoder inside (its own page, reviewed in effects/video_vocoder/review.md), and every effect inside rebuilt for lib's seven band channels and pyramid(); dev/probes/structure_diff.py finds only the audio callbacks' text changed in each; the showreel's page is unchanged. 0.1.17: Detail inside (its own page, reviewed in effects/detail/review.md), and every effect inside rebuilt for lib pyramid()'s prefix= and what= (Detail names three pyramids); dev/probes/structure_diff.py finds nothing changed but where operators sit in fifteen of them and only the pyramid shaders' comment text in Video Vocoder; the showreel's page gains Detail's take settings. 0.1.18: Chladni Sand inside (its own page, reviewed in effects/chladni_sand/review.md); no lib change, so the other effects are as in 0.1.17. The showreel's page gains Chladni Sand's take settings. 0.1.19: Pixel Art inside (its own page, reviewed in effects/pixel_art/review.md); no lib change, so the other effects are as in 0.1.18. Its take in the hero is at its defaults, so the page gains no take settings. 0.1.20: Lens Flare inside (its own page, reviewed in effects/lens_flare/review.md); no lib change, so the other effects are as in 0.1.19. Its take in the hero is at its defaults, so the page gains no take settings. 0.1.21: Words inside (its own page, reviewed in effects/words/review.md); a lib change (assert_idiomatic's P3 compares an operator parameter's path; optical_flow() can stop at 1/8) rebuilt every other effect, proven unchanged inside by dev/probes/structure_diff.py, so they are as in 0.1.20. Its take in the hero is at its defaults, so the page gains no take settings. 0.1.22: Scan Processor inside (its own page, reviewed in effects/scan_processor/review.md); no lib change, so the other effects are as in 0.1.21. Its take in the hero is at its defaults (Wobbulator), so the page gains no take settings. 0.1.23: Voxel Space inside (its own page, reviewed in effects/voxel_space/review.md); no lib change, so the other effects are as in 0.1.22. Its take in the hero is at its defaults (Fly), so the page gains no take settings. 0.1.24: Temporal Resonator inside (its own page, reviewed in effects/temporal_resonator/review.md). lib's audio_in() gains beat_period, so every effect inside was rebuilt with no version bump; dev/probes/structure_diff.py shows only each one's audio callbacks' text changed. Its take in the hero is at its defaults, so the page gains no take settings. 0.1.25: Caustics inside (its own page, reviewed in effects/caustics/review.md); no lib change, so the other effects are as in 0.1.24. Its take in the hero is at its defaults, so the page gains no take settings. 0.1.26: lib optical_flow() gains flow=False, so every effect inside was rebuilt. Caustics 0.1.1 and Temporal Resonator 0.1.1 leave out the optical flow they never read (lk8, lk4 and lum8p, 11 operators in all), and dev/probes/structure_diff.py shows nothing else changed but version strings. The page is unchanged. 0.1.27: Relight inside (its own page, reviewed in effects/relight/review.md); no lib change, so the other effects are as in 0.1.26. Its take in the hero is at its defaults, so the page gains no take settings. 0.1.28: Film Stock inside (its own page, reviewed in effects/film_stock/review.md); no lib change, so the other effects are as in 0.1.27. 0.1.29: Particles 0.4.0 inside, with the new Behavior Ribbons on its own page (reviewed in effects/particles/review.md); the showreel shows it at its default, Emit, so its page and take are unchanged. + Menu for View (discrete choices) and Flashing; Toggle for Labels and Full Screen; Int for Every and Display; Pulses for Open Window and Open Docs. Labels in Title Case; names are 10 characters or fewer (save-time check passes). The effects inside keep their own reviewed pages. 0.1.3 adds Fluid inside (its own page of parameters, reviewed in effects/fluid/review.md); the showreel's page is unchanged. 0.1.4 adds VHS + CRT inside (its own page, reviewed in effects/vhs_crt/review.md); the showreel's page is unchanged. 0.1.5: VHS + CRT 0.2.0 inside (its page gains Screen and the screens' settings, reviewed in effects/vhs_crt/review.md); the showreel's page is unchanged. 0.1.6: the effects inside are rebuilt for the clock fix (Blob Track 0.1.3, Datamosh 0.2.1, Fluid 0.1.1, Particles 0.3.2, Point Cloud 0.1.3, VHS + CRT 0.2.1); the showreel's page is unchanged. 0.1.7 adds Bent Cam inside (its own page, reviewed in effects/bent_cam/review.md), and every effect inside is rebuilt for the shared red guard's move into lib/glsl; the showreel's page is unchanged. 0.1.8: Bent Cam 0.1.1 inside (its page gains Beat Kick, reviewed in effects/bent_cam/review.md); the showreel's page is unchanged. 0.1.9: Slit-Scan inside (its own page, reviewed in effects/slit_scan/review.md); the showreel's page is unchanged. 0.1.10: Pixel Sort inside (its own page, reviewed in effects/pixel_sort/review.md); the showreel's page is unchanged. 0.1.11: Paint inside (its own page, reviewed in effects/paint/review.md), and every effect inside rebuilt for lib's new structure tensor; the showreel's page is unchanged. 0.1.13: Reaction-Diffusion inside (its own page, reviewed in effects/reaction_diffusion/review.md), and every effect inside rebuilt for lib's structure_tensor(ease=); the showreel's page is unchanged. 0.1.14: Pencil inside (its own page, reviewed in effects/pencil/review.md), and every effect inside rebuilt for lib's paper.glsl (moved from Paint unchanged); the showreel's page is unchanged. 0.1.15: Motion Magnify inside (its own page, reviewed in effects/motion_magnify/review.md); the showreel's page is unchanged. 0.1.16: Video Vocoder inside (its own page, reviewed in effects/video_vocoder/review.md), and every effect inside rebuilt for lib's seven band channels and pyramid(); dev/probes/structure_diff.py finds only the audio callbacks' text changed in each; the showreel's page is unchanged. 0.1.17: Detail inside (its own page, reviewed in effects/detail/review.md), and every effect inside rebuilt for lib pyramid()'s prefix= and what= (Detail names three pyramids); dev/probes/structure_diff.py finds nothing changed but where operators sit in fifteen of them and only the pyramid shaders' comment text in Video Vocoder; the showreel's page gains Detail's take settings. 0.1.18: Chladni Sand inside (its own page, reviewed in effects/chladni_sand/review.md); no lib change, so the other effects are as in 0.1.17. The showreel's page gains Chladni Sand's take settings. 0.1.19: Pixel Art inside (its own page, reviewed in effects/pixel_art/review.md); no lib change, so the other effects are as in 0.1.18. Its take in the hero is at its defaults, so the page gains no take settings. 0.1.20: Lens Flare inside (its own page, reviewed in effects/lens_flare/review.md); no lib change, so the other effects are as in 0.1.19. Its take in the hero is at its defaults, so the page gains no take settings. 0.1.21: Words inside (its own page, reviewed in effects/words/review.md); a lib change (assert_idiomatic's P3 compares an operator parameter's path; optical_flow() can stop at 1/8) rebuilt every other effect, proven unchanged inside by dev/probes/structure_diff.py, so they are as in 0.1.20. Its take in the hero is at its defaults, so the page gains no take settings. 0.1.22: Scan Processor inside (its own page, reviewed in effects/scan_processor/review.md); no lib change, so the other effects are as in 0.1.21. Its take in the hero is at its defaults (Wobbulator), so the page gains no take settings. 0.1.23: Voxel Space inside (its own page, reviewed in effects/voxel_space/review.md); no lib change, so the other effects are as in 0.1.22. Its take in the hero is at its defaults (Fly), so the page gains no take settings. 0.1.24: Temporal Resonator inside (its own page, reviewed in effects/temporal_resonator/review.md). lib's audio_in() gains beat_period, so every effect inside was rebuilt with no version bump; dev/probes/structure_diff.py shows only each one's audio callbacks' text changed. Its take in the hero is at its defaults, so the page gains no take settings. 0.1.25: Caustics inside (its own page, reviewed in effects/caustics/review.md); no lib change, so the other effects are as in 0.1.24. Its take in the hero is at its defaults, so the page gains no take settings. 0.1.26: lib optical_flow() gains flow=False, so every effect inside was rebuilt. Caustics 0.1.1 and Temporal Resonator 0.1.1 leave out the optical flow they never read (lk8, lk4 and lum8p, 11 operators in all), and dev/probes/structure_diff.py shows nothing else changed but version strings. The page is unchanged. 0.1.27: Relight inside (its own page, reviewed in effects/relight/review.md); no lib change, so the other effects are as in 0.1.26. Its take in the hero is at its defaults, so the page gains no take settings. 0.1.28: Film Stock inside (its own page, reviewed in effects/film_stock/review.md); no lib change, so the other effects are as in 0.1.27. 0.1.29: Particles 0.4.0 inside, with the new Behavior Ribbons on its own page (reviewed in effects/particles/review.md); the showreel shows it at its default, Emit, so its page and take are unchanged. 0.1.30: unchanged. Every effect inside is rebuilt for lib's new parts (warmth(), incompressible(), spectrum.glsl); dev/probes/structure_diff.py finds only shader text changed (the headers of Fluid's divergence, pressure and velocity, and Lens Flare's burst with the CIE fit now from lib). - [x] P5 Inner operators follow the COMP's parameters through expressions (`parent().par.Amount`), not scripts copying values; values only the inside needs are kept off the pages (`ipar`). [Internal Parameters](https://docs.derivative.ca/Internal_Parameters) - The page drives the inside through expressions: view's index (parent().par.View, Every through the cycle CHOP's step and fade), label's uniforms (parent().par.Labels), and every embedded effect's Flashing (parent().par.Flashing). No script copies values. The names, the 2 s bar, the 2 s minimum gap and the 1 s fade live in the builder and src/, off the page. 0.1.3: unchanged; Fluid's Flashing follows the showreel's like the others'. 0.1.4: unchanged; VHS + CRT's Flashing follows the showreel's like the others'. 0.1.5: unchanged. 0.1.7: unchanged; Bent Cam's Flashing follows the showreel's like the others'. 0.1.8: unchanged. 0.1.9: each effect's Flashing follows the showreel's through an expression that holds Up to 3 per second at Calm while Cycle fades and for a second after (cycle's new since channel); still no script copying values. 0.1.10: unchanged; Pixel Sort's Flashing follows the showreel's through the same expression as the others'. 0.1.11: unchanged; Paint has no Flashing parameter, so the showreel's Flashing expression has nothing to drive in it, as in Point Cloud. 0.1.13: unchanged; Reaction-Diffusion's Flashing follows the showreel's through the same expression as the others'. 0.1.14: unchanged; Pencil has no Flashing setting (it makes no flash of its own, its review A2). 0.1.15: unchanged; Motion Magnify has no Flashing setting (it makes no flash of its own, its review A2). 0.1.16: unchanged; Video Vocoder's Flashing follows the showreel's through the same expression as the others'. 0.1.17: unchanged; Detail's Flashing follows the showreel's through the same expression as the others'. 0.1.18: unchanged; Chladni Sand's Flashing follows the showreel's through the same expression as the others'. 0.1.19: unchanged; Pixel Art's Flashing follows the showreel's through the same expression as the others'. 0.1.20: unchanged; Lens Flare's Flashing follows the showreel's through the same expression as the others'. 0.1.21: unchanged; Words' Flashing follows the showreel's through the same expression as the others'. 0.1.22: unchanged; Scan Processor's Flashing follows the showreel's through the same expression as the others'. 0.1.23: unchanged; Voxel Space's Flashing follows the showreel's through the same expression as the others'. 0.1.24: unchanged; Temporal Resonator has no Flashing setting (nothing in it flashes, its review A2), so the showreel's Flashing passes it by, as it does Motion Magnify. 0.1.25: unchanged; Caustics has no Flashing setting (nothing in it flashes, its review A2), so the showreel's Flashing passes it by. 0.1.27: unchanged; Relight's Flashing follows the showreel's through the same expression as the others'. + The page drives the inside through expressions: view's index (parent().par.View, Every through the cycle CHOP's step and fade), label's uniforms (parent().par.Labels), and every embedded effect's Flashing (parent().par.Flashing). No script copies values. The names, the 2 s bar, the 2 s minimum gap and the 1 s fade live in the builder and src/, off the page. 0.1.3: unchanged; Fluid's Flashing follows the showreel's like the others'. 0.1.4: unchanged; VHS + CRT's Flashing follows the showreel's like the others'. 0.1.5: unchanged. 0.1.7: unchanged; Bent Cam's Flashing follows the showreel's like the others'. 0.1.8: unchanged. 0.1.9: each effect's Flashing follows the showreel's through an expression that holds Up to 3 per second at Calm while Cycle fades and for a second after (cycle's new since channel); still no script copying values. 0.1.10: unchanged; Pixel Sort's Flashing follows the showreel's through the same expression as the others'. 0.1.11: unchanged; Paint has no Flashing parameter, so the showreel's Flashing expression has nothing to drive in it, as in Point Cloud. 0.1.13: unchanged; Reaction-Diffusion's Flashing follows the showreel's through the same expression as the others'. 0.1.14: unchanged; Pencil has no Flashing setting (it makes no flash of its own, its review A2). 0.1.15: unchanged; Motion Magnify has no Flashing setting (it makes no flash of its own, its review A2). 0.1.16: unchanged; Video Vocoder's Flashing follows the showreel's through the same expression as the others'. 0.1.17: unchanged; Detail's Flashing follows the showreel's through the same expression as the others'. 0.1.18: unchanged; Chladni Sand's Flashing follows the showreel's through the same expression as the others'. 0.1.19: unchanged; Pixel Art's Flashing follows the showreel's through the same expression as the others'. 0.1.20: unchanged; Lens Flare's Flashing follows the showreel's through the same expression as the others'. 0.1.21: unchanged; Words' Flashing follows the showreel's through the same expression as the others'. 0.1.22: unchanged; Scan Processor's Flashing follows the showreel's through the same expression as the others'. 0.1.23: unchanged; Voxel Space's Flashing follows the showreel's through the same expression as the others'. 0.1.24: unchanged; Temporal Resonator has no Flashing setting (nothing in it flashes, its review A2), so the showreel's Flashing passes it by, as it does Motion Magnify. 0.1.25: unchanged; Caustics has no Flashing setting (nothing in it flashes, its review A2), so the showreel's Flashing passes it by. 0.1.27: unchanged; Relight's Flashing follows the showreel's through the same expression as the others'. 0.1.30: unchanged. Every effect inside is rebuilt for lib's new parts (warmth(), incompressible(), spectrum.glsl); dev/probes/structure_diff.py finds only shader text changed (the headers of Fluid's divergence, pressure and velocity, and Lens Flare's burst with the CIE fit now from lib). - [x] P6 Related parameters are grouped with `startSection`; parameters that don't apply in the current mode are grayed out with `enableExpr`. - Sections: View, Every and Labels; then Flashing; then the window. Every is grayed out unless View is Cycle (enableExpr); Labels applies to every view. 0.1.3: unchanged. 0.1.4: unchanged. 0.1.5: unchanged. 0.1.7: unchanged. 0.1.8: unchanged. 0.1.9: unchanged. 0.1.10: unchanged. 0.1.11: unchanged. 0.1.12: unchanged. 0.1.13: unchanged. 0.1.14: unchanged. 0.1.15: unchanged. 0.1.16: unchanged. 0.1.17: unchanged. 0.1.18: unchanged. 0.1.19: unchanged. 0.1.20: unchanged. 0.1.21: unchanged. 0.1.22: unchanged. 0.1.23: unchanged. 0.1.24: unchanged. 0.1.25: unchanged. 0.1.26: unchanged. 0.1.27: unchanged. 0.1.29: unchanged. + Sections: View, Every and Labels; then Flashing; then the window. Every is grayed out unless View is Cycle (enableExpr); Labels applies to every view. 0.1.3: unchanged. 0.1.4: unchanged. 0.1.5: unchanged. 0.1.7: unchanged. 0.1.8: unchanged. 0.1.9: unchanged. 0.1.10: unchanged. 0.1.11: unchanged. 0.1.12: unchanged. 0.1.13: unchanged. 0.1.14: unchanged. 0.1.15: unchanged. 0.1.16: unchanged. 0.1.17: unchanged. 0.1.18: unchanged. 0.1.19: unchanged. 0.1.20: unchanged. 0.1.21: unchanged. 0.1.22: unchanged. 0.1.23: unchanged. 0.1.24: unchanged. 0.1.25: unchanged. 0.1.26: unchanged. 0.1.27: unchanged. 0.1.29: unchanged. 0.1.30: unchanged. Every effect inside is rebuilt for lib's new parts (warmth(), incompressible(), spectrum.glsl); dev/probes/structure_diff.py finds only shader text changed (the headers of Fluid's divergence, pressure and velocity, and Lens Flare's burst with the CIE fit now from lib). - [x] I4 The audio input goes through `audio_in()`, which reads it by channel name with fallbacks and takes a pulse from its count channel when one exists, because a one-frame pulse is lost on any frame nothing pulls the CHOP; the effect reads the counts (`beat_count`, `bar_count`...) for logic, not `beat`. [Cook](https://docs.derivative.ca/Cook), [Logic CHOP](https://docs.derivative.ca/Logic_CHOP) - Input 2 reaches each embedded effect unchanged (each goes through its own audio_in()); the showreel's own audio_in() feeds only cycle, which reads the bar_count and source channels, never the one-frame beat pulse. 0.1.1: the effects inside and the showreel's own audio_in() take raw audio's beats from the beat clock (lib/audio_callbacks.py, #4652), so Cycle's bars on raw audio come every 4 locked beats. 0.1.3: Fluid reads input 2 through its own audio_in() too (beat_count and drop_count through its flash gate). 0.1.4: VHS + CRT reads input 2 through its own audio_in() (beat_count, bar_count and drop_count through its flash gate). 0.1.5: VHS + CRT 0.2.0 reads input 2 as 0.1.0 did (its e-ink refresh comes off the same gated bursts). 0.1.7: Bent Cam reads input 2 through its own audio_in() (beat_count, bar_count and drop_count, through its flash gate, and build as a level). 0.1.8: Bent Cam 0.1.1 reads input 2 as 0.1.0 did, its beat count now also driving the crackle. 0.1.9: Slit-Scan reads input 2 through its own audio_in() (beat_count, bar_count, drop_count through its gate, level and build as levels); cycle still reads only bar_count and source. 0.1.10: Pixel Sort reads input 2 through its own audio_in() (beat_count, bar_count, drop_count through its gate, level and build as levels); cycle still reads only bar_count and source. 0.1.11: Paint reads input 2 through its own audio_in() (beat_count and drop_count as steps, level and build as levels); cycle still reads only bar_count and source. 0.1.13: Reaction-Diffusion reads input 2 through its own audio_in() (beat_count and drop_count as one, through its flash gate; level as a level); cycle still reads only bar_count and source. 0.1.14: Pencil reads input 2 through its own audio_in() (beat_count and drop_count as counts, build and level as levels); cycle still reads only bar_count and source. 0.1.15: Motion Magnify reads input 2 through its own audio_in() (beat_count and drop_count as steps, build and level as levels); cycle still reads only bar_count and source. 0.1.16: Video Vocoder reads input 2 through its own audio_in() (its seven bands as levels, through its own flash gate); audio_in() now also gives the bands, which cycle doesn't read: it still reads only bar_count and source. 0.1.17: Detail reads input 2 through its own audio_in() (the beat count through its own flash gate, the level against its usual level); cycle still reads only bar_count and source. 0.1.18: Chladni Sand reads input 2 through its own audio_in() (its seven bands against their usual level, the beat count through its own flash gate); cycle still reads only bar_count and source. 0.1.19: Pixel Art reads input 2 through its own audio_in() (the beat count through its own flash gate, the loudness against its usual level); cycle still reads only bar_count and source. 0.1.20: Lens Flare reads input 2 through its own audio_in() (the beat count through its own flash gate, the loudness against its usual level); cycle still reads only bar_count and source. 0.1.21: Words reads input 2 through its own audio_in() (the beat count through its own flash gate, the loudness against its usual level); cycle still reads only bar_count and source. 0.1.22: Scan Processor reads input 2 through its own audio_in() (the beat count through its own flash gate, the seven bands smoothed and against their usual level); cycle still reads only bar_count and source. 0.1.23: Voxel Space reads input 2 through its own audio_in() (the beat count through its own flash gate for the dip, the low bands smoothed for the lift, the beats counted for the pace); cycle still reads only bar_count and source. 0.1.24: Temporal Resonator reads input 2 through its own audio_in() (beat_period for its strings' length, beat_phase for its plucks, the level for Loudness); cycle still reads only bar_count and source. 0.1.25: Caustics reads input 2 through its own audio_in() (beat_count for its drops, the level for its waves); cycle still reads only bar_count and source. 0.1.27: Relight reads input 2 through its own audio_in() (beat_count through its own flash gate for its flares, the level for its brightness); cycle still reads only bar_count and source. + Input 2 reaches each embedded effect unchanged (each goes through its own audio_in()); the showreel's own audio_in() feeds only cycle, which reads the bar_count and source channels, never the one-frame beat pulse. 0.1.1: the effects inside and the showreel's own audio_in() take raw audio's beats from the beat clock (lib/audio_callbacks.py, #4652), so Cycle's bars on raw audio come every 4 locked beats. 0.1.3: Fluid reads input 2 through its own audio_in() too (beat_count and drop_count through its flash gate). 0.1.4: VHS + CRT reads input 2 through its own audio_in() (beat_count, bar_count and drop_count through its flash gate). 0.1.5: VHS + CRT 0.2.0 reads input 2 as 0.1.0 did (its e-ink refresh comes off the same gated bursts). 0.1.7: Bent Cam reads input 2 through its own audio_in() (beat_count, bar_count and drop_count, through its flash gate, and build as a level). 0.1.8: Bent Cam 0.1.1 reads input 2 as 0.1.0 did, its beat count now also driving the crackle. 0.1.9: Slit-Scan reads input 2 through its own audio_in() (beat_count, bar_count, drop_count through its gate, level and build as levels); cycle still reads only bar_count and source. 0.1.10: Pixel Sort reads input 2 through its own audio_in() (beat_count, bar_count, drop_count through its gate, level and build as levels); cycle still reads only bar_count and source. 0.1.11: Paint reads input 2 through its own audio_in() (beat_count and drop_count as steps, level and build as levels); cycle still reads only bar_count and source. 0.1.13: Reaction-Diffusion reads input 2 through its own audio_in() (beat_count and drop_count as one, through its flash gate; level as a level); cycle still reads only bar_count and source. 0.1.14: Pencil reads input 2 through its own audio_in() (beat_count and drop_count as counts, build and level as levels); cycle still reads only bar_count and source. 0.1.15: Motion Magnify reads input 2 through its own audio_in() (beat_count and drop_count as steps, build and level as levels); cycle still reads only bar_count and source. 0.1.16: Video Vocoder reads input 2 through its own audio_in() (its seven bands as levels, through its own flash gate); audio_in() now also gives the bands, which cycle doesn't read: it still reads only bar_count and source. 0.1.17: Detail reads input 2 through its own audio_in() (the beat count through its own flash gate, the level against its usual level); cycle still reads only bar_count and source. 0.1.18: Chladni Sand reads input 2 through its own audio_in() (its seven bands against their usual level, the beat count through its own flash gate); cycle still reads only bar_count and source. 0.1.19: Pixel Art reads input 2 through its own audio_in() (the beat count through its own flash gate, the loudness against its usual level); cycle still reads only bar_count and source. 0.1.20: Lens Flare reads input 2 through its own audio_in() (the beat count through its own flash gate, the loudness against its usual level); cycle still reads only bar_count and source. 0.1.21: Words reads input 2 through its own audio_in() (the beat count through its own flash gate, the loudness against its usual level); cycle still reads only bar_count and source. 0.1.22: Scan Processor reads input 2 through its own audio_in() (the beat count through its own flash gate, the seven bands smoothed and against their usual level); cycle still reads only bar_count and source. 0.1.23: Voxel Space reads input 2 through its own audio_in() (the beat count through its own flash gate for the dip, the low bands smoothed for the lift, the beats counted for the pace); cycle still reads only bar_count and source. 0.1.24: Temporal Resonator reads input 2 through its own audio_in() (beat_period for its strings' length, beat_phase for its plucks, the level for Loudness); cycle still reads only bar_count and source. 0.1.25: Caustics reads input 2 through its own audio_in() (beat_count for its drops, the level for its waves); cycle still reads only bar_count and source. 0.1.27: Relight reads input 2 through its own audio_in() (beat_count through its own flash gate for its flares, the level for its brightness); cycle still reads only bar_count and source. 0.1.30: unchanged. Every effect inside is rebuilt for lib's new parts (warmth(), incompressible(), spectrum.glsl); dev/probes/structure_diff.py finds only shader text changed (the headers of Fluid's divergence, pressure and velocity, and Lens Flare's burst with the CIE fit now from lib). - [x] E2 Only capitalized members are promoted; state lives on `self` and the extension can be reinitialized at any time; no module globals; `onInitTD`/`onDestroyTD`, the 2025 names. [Release Notes 2025](https://docs.derivative.ca/Release_Notes/2025.30000) - No extension in the showreel; the effects inside have none either. 0.1.3: Fluid has none either. 0.1.4: VHS + CRT has none either. 0.1.5: VHS + CRT has none either. 0.1.7: Bent Cam has none either. 0.1.8: unchanged. 0.1.9: Slit-Scan has none either. 0.1.10: Pixel Sort has none either. 0.1.11: Paint has none either. 0.1.13: Reaction-Diffusion has none either. 0.1.14: Pencil has none either. 0.1.15: Motion Magnify has none either. 0.1.16: Video Vocoder has none either. 0.1.17: Detail has none either. 0.1.18: Chladni Sand has none either. 0.1.19: Pixel Art has none either. 0.1.20: Lens Flare has none either. 0.1.21: Words has none either. 0.1.22: Scan Processor has none either. 0.1.23: Voxel Space has none either. 0.1.24: Temporal Resonator has none either. 0.1.25: Caustics has none either. 0.1.27: Relight has none either. + No extension in the showreel; the effects inside have none either. 0.1.3: Fluid has none either. 0.1.4: VHS + CRT has none either. 0.1.5: VHS + CRT has none either. 0.1.7: Bent Cam has none either. 0.1.8: unchanged. 0.1.9: Slit-Scan has none either. 0.1.10: Pixel Sort has none either. 0.1.11: Paint has none either. 0.1.13: Reaction-Diffusion has none either. 0.1.14: Pencil has none either. 0.1.15: Motion Magnify has none either. 0.1.16: Video Vocoder has none either. 0.1.17: Detail has none either. 0.1.18: Chladni Sand has none either. 0.1.19: Pixel Art has none either. 0.1.20: Lens Flare has none either. 0.1.21: Words has none either. 0.1.22: Scan Processor has none either. 0.1.23: Voxel Space has none either. 0.1.24: Temporal Resonator has none either. 0.1.25: Caustics has none either. 0.1.27: Relight has none either. 0.1.30: unchanged. Every effect inside is rebuilt for lib's new parts (warmth(), incompressible(), spectrum.glsl); dev/probes/structure_diff.py finds only shader text changed (the headers of Fluid's divergence, pressure and velocity, and Lens Flare's burst with the CIE fit now from lib). - [x] C4 Motion and timing follow seconds (`absTime.seconds`, `me.time.seconds`), not frame counts, so dropped frames don't change the speed. [AbsTime Class](https://docs.derivative.ca/AbsTime_Class) - Cycle's timing is in seconds: src/cycle.py reads absTime.seconds for the 2 s bar without music, the 2 s minimum gap and the 1 s fade, and counts input 2's bars; dev/probes/showreel_cycle.py measured moves exactly 2.0 s apart with nothing wired and under a 2-bars-a-second beat at Every 1. Nothing counts frames. 0.1.3: unchanged; Fluid steps its simulation by absTime.stepSeconds (its review, C4). 0.1.4: unchanged; VHS + CRT times everything by absTime.seconds (its review, C4). 0.1.5: unchanged; VHS + CRT 0.2.0 times its screens by absTime (its review, C4; e-ink and plasma count in Python). 0.1.6: the effects inside hand their shaders lib clock_seconds() (wrapped every 16384 s, board #4681); cycle.py reads absTime.seconds in Python (double precision), the clock CHOP only makes it cook. 0.1.7: unchanged; Bent Cam times everything by absTime.seconds, its camera frames counted in Python (its review, C4). 0.1.8: unchanged. 0.1.9: cycle's since channel is absTime.seconds minus the last move, in seconds; Slit-Scan times everything by seconds (its review, C4). 0.1.10: unchanged; Pixel Sort times everything by seconds (its review, C4). 0.1.11: unchanged; Paint times everything by seconds (its review, C4). 0.1.13: unchanged; Reaction-Diffusion steps its model and eases by the frame step (its review, C4). 0.1.14: unchanged; Pencil eases and carries its strokes by the frame step (its review, C4). 0.1.15: unchanged; Motion Magnify carries, leaks and eases its displacement by the frame step (its review, C4). 0.1.16: unchanged; Video Vocoder times its bands in seconds (Lag, Filter and Trigger CHOPs, its review, C4). 0.1.17: unchanged; Detail times its jump and smoothing in seconds (Trigger and Filter CHOPs, its review, C4). 0.1.18: unchanged; Chladni Sand times its strikes and smoothing in seconds and moves its sand by the frame's step (its review, C4). 0.1.19: unchanged; Pixel Art times its mosaic, easing, fade-in and annealing in seconds (its review, C4). 0.1.20: unchanged; Lens Flare times its kick and its easing in seconds (its review, C4). 0.1.21: unchanged; Words times its fades, holds, lines and pump in seconds (its review, C4). 0.1.22: unchanged; Scan Processor times its generators' creep by the frame's step, its strikes and smoothing in seconds (its review, C4). 0.1.23: unchanged; Voxel Space moves its camera and eases its land by the frame's step, its dip and smoothing in seconds (its review, C4). 0.1.24: unchanged; Temporal Resonator's delay, glide and background step are in seconds and it finds its slots by their stamped times (its review, C4). 0.1.25: unchanged; Caustics steps its water by the frame's step and moves its swells by a clock in seconds (its review, C4). 0.1.27: unchanged; Relight moves its lights by a clock in seconds and fades its flares over 0.6 s (its review, C4). 0.1.28: unchanged; Film Stock weaves and flickers by a clock in seconds and counts the film's frames from the seconds in Python (its review, C4). + Cycle's timing is in seconds: src/cycle.py reads absTime.seconds for the 2 s bar without music, the 2 s minimum gap and the 1 s fade, and counts input 2's bars; dev/probes/showreel_cycle.py measured moves exactly 2.0 s apart with nothing wired and under a 2-bars-a-second beat at Every 1. Nothing counts frames. 0.1.3: unchanged; Fluid steps its simulation by absTime.stepSeconds (its review, C4). 0.1.4: unchanged; VHS + CRT times everything by absTime.seconds (its review, C4). 0.1.5: unchanged; VHS + CRT 0.2.0 times its screens by absTime (its review, C4; e-ink and plasma count in Python). 0.1.6: the effects inside hand their shaders lib clock_seconds() (wrapped every 16384 s, board #4681); cycle.py reads absTime.seconds in Python (double precision), the clock CHOP only makes it cook. 0.1.7: unchanged; Bent Cam times everything by absTime.seconds, its camera frames counted in Python (its review, C4). 0.1.8: unchanged. 0.1.9: cycle's since channel is absTime.seconds minus the last move, in seconds; Slit-Scan times everything by seconds (its review, C4). 0.1.10: unchanged; Pixel Sort times everything by seconds (its review, C4). 0.1.11: unchanged; Paint times everything by seconds (its review, C4). 0.1.13: unchanged; Reaction-Diffusion steps its model and eases by the frame step (its review, C4). 0.1.14: unchanged; Pencil eases and carries its strokes by the frame step (its review, C4). 0.1.15: unchanged; Motion Magnify carries, leaks and eases its displacement by the frame step (its review, C4). 0.1.16: unchanged; Video Vocoder times its bands in seconds (Lag, Filter and Trigger CHOPs, its review, C4). 0.1.17: unchanged; Detail times its jump and smoothing in seconds (Trigger and Filter CHOPs, its review, C4). 0.1.18: unchanged; Chladni Sand times its strikes and smoothing in seconds and moves its sand by the frame's step (its review, C4). 0.1.19: unchanged; Pixel Art times its mosaic, easing, fade-in and annealing in seconds (its review, C4). 0.1.20: unchanged; Lens Flare times its kick and its easing in seconds (its review, C4). 0.1.21: unchanged; Words times its fades, holds, lines and pump in seconds (its review, C4). 0.1.22: unchanged; Scan Processor times its generators' creep by the frame's step, its strikes and smoothing in seconds (its review, C4). 0.1.23: unchanged; Voxel Space moves its camera and eases its land by the frame's step, its dip and smoothing in seconds (its review, C4). 0.1.24: unchanged; Temporal Resonator's delay, glide and background step are in seconds and it finds its slots by their stamped times (its review, C4). 0.1.25: unchanged; Caustics steps its water by the frame's step and moves its swells by a clock in seconds (its review, C4). 0.1.27: unchanged; Relight moves its lights by a clock in seconds and fades its flares over 0.6 s (its review, C4). 0.1.28: unchanged; Film Stock weaves and flickers by a clock in seconds and counts the film's frames from the seconds in Python (its review, C4). 0.1.30: unchanged. Every effect inside is rebuilt for lib's new parts (warmth(), incompressible(), spectrum.glsl); dev/probes/structure_diff.py finds only shader text changed (the headers of Fluid's divergence, pressure and velocity, and Lens Flare's burst with the CIE fit now from lib). - [x] C5 Nothing cooks while the output is unused: checked with `dev/idle.py` (nothing downstream, the node out of view; input 2 unwired and wired to a changing beat), or in the Performance Monitor. [Performance Monitor](https://docs.derivative.ca/Performance_Monitor_Dialog) - dev/idle.py on the 0.1.0 build with ascii, blob_track, datamosh, glitch, particles and point_cloud inside and the dancer's depth carried with her clip (node viewer off, nothing downstream): 0 of 513 operators cooked over 60 frames unused, with nothing wired and with the clip, its depth on input 3 and a changing beat wired in; the control cooked 56 and 58. The cycle CHOP and its clock cook only while View is Cycle (the view reads them only then); the blending Switch TOP cooks only the inputs it shows, one effect, or two during a fade (dev/probes/showreel_fade.py logged which cooked every frame); the two built-in clips show the frame the clock picks, which costs nothing while nothing pulls them. dev/idle.py on the 0.1.1 build (ascii 0.2.0, blob_track 0.1.2, datamosh 0.2.0, glitch 0.1.1, particles 0.3.0, point_cloud 0.1.1 inside): 0 of 531 operators cooked over 60 frames unused, unwired and wired; control 57 and 59. dev/idle.py on the 0.1.2 build (Particles 0.3.1, Point Cloud 0.1.2 and the new dancer depth inside): 0 of 531 operators cooked over 60 frames unused, unwired and wired; control 57 and 59. dev/idle.py on the 0.1.3 build (Fluid 0.1.0 added; every effect rebuilt for the lib change): 0 of 643 operators cooked over 60 frames unused, unwired and wired; control 57 and 59. dev/idle.py on the 0.1.4 build (VHS + CRT 0.1.0 added): 0 of 706 operators cooked over 60 frames unused, unwired and wired; control 57 and 59. dev/idle.py on the 0.1.5 build (VHS + CRT 0.2.0 inside, every effect rebuilt for the lib change): 0 of 770 operators cooked over 60 frames unused, unwired and wired; control 57 and 59. dev/idle.py on the 0.1.7 build (Bent Cam 0.1.0 added, every effect rebuilt for the lib change): 0 of 840 operators cooked over 60 frames unused, unwired and wired; control 57 and 59. dev/idle.py on the 0.1.8 build: 0 of 840 operators cooked over 60 frames unused, unwired and wired; control 57 and 59. 0.1.9 (ten effects): dev/idle.py, 0 of 951 operators cooked over 60 frames unused, unwired and wired; control 57 and 59. The effects' Flashing expressions read cycle only while View is Cycle, and only when an effect evaluates its parameters, which it does only while it cooks. 0.1.10 (eleven effects): dev/idle.py, 0 of 1012 operators cooked over 60 frames unused, unwired and wired; control 57 and 59. 0.1.11 (twelve effects): dev/idle.py, 0 of 1092 operators cooked over 60 frames unused, unwired and wired; control 57 and 59. dev/idle.py on the 0.1.12 build: 0 of 3752 operators (the Network Boxes' insides included) cooked over 60 frames unused, unwired and wired; control 57 and 59. 0.1.13 (thirteen effects): dev/idle.py, 0 of 4071 operators cooked over 60 frames unused, unwired and wired; control 57 and 59. 0.1.14 (fourteen effects): dev/idle.py, 0 of 4442 operators cooked over 60 frames unused, unwired and wired; control 57 and 59. 0.1.15 (fifteen effects): dev/idle.py, 0 of 4713 operators cooked over 60 frames unused, unwired and wired; control 57 and 59. 0.1.16 (sixteen effects): dev/idle.py, 0 of 4951 operators cooked over 60 frames unused, unwired and wired; control 107 and 59. 0.1.17 (seventeen effects): dev/idle.py, 0 of 5263 operators cooked over 60 frames unused, unwired and wired; control 57 and 59. 0.1.18 (eighteen effects): dev/idle.py, 0 of 5472 operators cooked over 60 frames unused, unwired and wired; control 57 and 59. 0.1.19 (nineteen effects): dev/idle.py, 0 of 5727 operators cooked over 60 frames unused, unwired and wired; control 107 and 59. 0.1.20 (twenty effects): dev/idle.py, 0 of 6014 operators cooked over 60 frames unused, unwired and wired; control 57 and 59. 0.1.21 (twenty-one effects): dev/idle.py, 0 of 6388 operators cooked over 60 frames unused, unwired and wired; control 57 and 110. 0.1.22 (twenty-two effects): dev/idle.py, 0 of 6572 operators cooked over 60 frames unused, unwired and wired; control 57 and 59. 0.1.23 (twenty-three effects): dev/idle.py, 0 of 6832 operators cooked over 60 frames unused, unwired and wired; control 57 and 59. 0.1.24 (twenty-four effects): dev/idle.py, 0 of 7067 operators cooked over 60 frames unused, unwired and wired; control 57 and 110. 0.1.25 (twenty-five effects): dev/idle.py, 0 of 7303 operators cooked over 60 frames unused, unwired and wired; control 57 and 110. 0.1.26 (twenty-five effects): dev/idle.py, 0 of 7292 operators cooked over 60 frames unused, unwired and wired; control 107 and 110. 0.1.27 (twenty-six effects): dev/idle.py, 0 of 7553 operators cooked over 60 frames unused, unwired and wired; control 57 and 110. 0.1.28 (twenty-seven effects): dev/idle.py, 0 of 7904 operators cooked over 60 frames unused, unwired and wired; control 57 and 110. 0.1.29 (twenty-seven effects, Particles 0.4.0 with middle_in): dev/idle.py, 0 of 7905 operators cooked over 60 frames unused, unwired and wired; control 57 and 110. + dev/idle.py on the 0.1.0 build with ascii, blob_track, datamosh, glitch, particles and point_cloud inside and the dancer's depth carried with her clip (node viewer off, nothing downstream): 0 of 513 operators cooked over 60 frames unused, with nothing wired and with the clip, its depth on input 3 and a changing beat wired in; the control cooked 56 and 58. The cycle CHOP and its clock cook only while View is Cycle (the view reads them only then); the blending Switch TOP cooks only the inputs it shows, one effect, or two during a fade (dev/probes/showreel_fade.py logged which cooked every frame); the two built-in clips show the frame the clock picks, which costs nothing while nothing pulls them. dev/idle.py on the 0.1.1 build (ascii 0.2.0, blob_track 0.1.2, datamosh 0.2.0, glitch 0.1.1, particles 0.3.0, point_cloud 0.1.1 inside): 0 of 531 operators cooked over 60 frames unused, unwired and wired; control 57 and 59. dev/idle.py on the 0.1.2 build (Particles 0.3.1, Point Cloud 0.1.2 and the new dancer depth inside): 0 of 531 operators cooked over 60 frames unused, unwired and wired; control 57 and 59. dev/idle.py on the 0.1.3 build (Fluid 0.1.0 added; every effect rebuilt for the lib change): 0 of 643 operators cooked over 60 frames unused, unwired and wired; control 57 and 59. dev/idle.py on the 0.1.4 build (VHS + CRT 0.1.0 added): 0 of 706 operators cooked over 60 frames unused, unwired and wired; control 57 and 59. dev/idle.py on the 0.1.5 build (VHS + CRT 0.2.0 inside, every effect rebuilt for the lib change): 0 of 770 operators cooked over 60 frames unused, unwired and wired; control 57 and 59. dev/idle.py on the 0.1.7 build (Bent Cam 0.1.0 added, every effect rebuilt for the lib change): 0 of 840 operators cooked over 60 frames unused, unwired and wired; control 57 and 59. dev/idle.py on the 0.1.8 build: 0 of 840 operators cooked over 60 frames unused, unwired and wired; control 57 and 59. 0.1.9 (ten effects): dev/idle.py, 0 of 951 operators cooked over 60 frames unused, unwired and wired; control 57 and 59. The effects' Flashing expressions read cycle only while View is Cycle, and only when an effect evaluates its parameters, which it does only while it cooks. 0.1.10 (eleven effects): dev/idle.py, 0 of 1012 operators cooked over 60 frames unused, unwired and wired; control 57 and 59. 0.1.11 (twelve effects): dev/idle.py, 0 of 1092 operators cooked over 60 frames unused, unwired and wired; control 57 and 59. dev/idle.py on the 0.1.12 build: 0 of 3752 operators (the Network Boxes' insides included) cooked over 60 frames unused, unwired and wired; control 57 and 59. 0.1.13 (thirteen effects): dev/idle.py, 0 of 4071 operators cooked over 60 frames unused, unwired and wired; control 57 and 59. 0.1.14 (fourteen effects): dev/idle.py, 0 of 4442 operators cooked over 60 frames unused, unwired and wired; control 57 and 59. 0.1.15 (fifteen effects): dev/idle.py, 0 of 4713 operators cooked over 60 frames unused, unwired and wired; control 57 and 59. 0.1.16 (sixteen effects): dev/idle.py, 0 of 4951 operators cooked over 60 frames unused, unwired and wired; control 107 and 59. 0.1.17 (seventeen effects): dev/idle.py, 0 of 5263 operators cooked over 60 frames unused, unwired and wired; control 57 and 59. 0.1.18 (eighteen effects): dev/idle.py, 0 of 5472 operators cooked over 60 frames unused, unwired and wired; control 57 and 59. 0.1.19 (nineteen effects): dev/idle.py, 0 of 5727 operators cooked over 60 frames unused, unwired and wired; control 107 and 59. 0.1.20 (twenty effects): dev/idle.py, 0 of 6014 operators cooked over 60 frames unused, unwired and wired; control 57 and 59. 0.1.21 (twenty-one effects): dev/idle.py, 0 of 6388 operators cooked over 60 frames unused, unwired and wired; control 57 and 110. 0.1.22 (twenty-two effects): dev/idle.py, 0 of 6572 operators cooked over 60 frames unused, unwired and wired; control 57 and 59. 0.1.23 (twenty-three effects): dev/idle.py, 0 of 6832 operators cooked over 60 frames unused, unwired and wired; control 57 and 59. 0.1.24 (twenty-four effects): dev/idle.py, 0 of 7067 operators cooked over 60 frames unused, unwired and wired; control 57 and 110. 0.1.25 (twenty-five effects): dev/idle.py, 0 of 7303 operators cooked over 60 frames unused, unwired and wired; control 57 and 110. 0.1.26 (twenty-five effects): dev/idle.py, 0 of 7292 operators cooked over 60 frames unused, unwired and wired; control 107 and 110. 0.1.27 (twenty-six effects): dev/idle.py, 0 of 7553 operators cooked over 60 frames unused, unwired and wired; control 57 and 110. 0.1.28 (twenty-seven effects): dev/idle.py, 0 of 7904 operators cooked over 60 frames unused, unwired and wired; control 57 and 110. 0.1.29 (twenty-seven effects, Particles 0.4.0 with middle_in): dev/idle.py, 0 of 7905 operators cooked over 60 frames unused, unwired and wired; control 57 and 110. 0.1.30: dev/idle.py on the 0.1.30 build: 0 of 7905 operators cooked over 60 frames unused, unwired and wired; the controls cooked 57 and 110. - [x] C6 Pixel formats are 8-bit unless precision is needed, and 32-bit float for data textures (a CHOP to TOP feeding a shader). - view and label are 8-bit (the pictures); the Null TOPs pass each effect's own format; no data textures of its own (the effects inside keep their reviewed formats). 0.1.3: unchanged; Fluid keeps its reviewed formats. 0.1.4: unchanged; VHS + CRT keeps its reviewed formats. 0.1.5: unchanged; VHS + CRT keeps its reviewed formats. 0.1.7: unchanged; Bent Cam keeps its reviewed formats. 0.1.8: unchanged. 0.1.9: unchanged; Slit-Scan keeps its reviewed formats. 0.1.10: unchanged; Pixel Sort keeps its reviewed formats. 0.1.11: unchanged; Paint keeps its reviewed formats. 0.1.13: unchanged; Reaction-Diffusion keeps its reviewed formats (its pattern in 32-bit float). 0.1.14: unchanged; Pencil keeps its reviewed formats (its strokes' offsets in 32-bit float). 0.1.15: unchanged; Motion Magnify keeps its reviewed formats (its displacement in 32-bit float). 0.1.16: unchanged; Video Vocoder keeps its reviewed formats (its pyramid and passes in 32-bit float). 0.1.17: unchanged; Detail keeps its reviewed formats (its pyramids and passes in 32-bit float). 0.1.18: unchanged; Chladni Sand keeps its reviewed formats (its plate and sand in 32-bit float). 0.1.19: unchanged; Pixel Art keeps its reviewed formats (its grid's data textures in 32-bit float). 0.1.20: unchanged; Lens Flare keeps its reviewed formats (its tiles, lights, ghost table and the starburst's transform in 32-bit float). 0.1.21: unchanged; Words keeps its reviewed formats (its word tables, cells, field and slots in 32-bit float). 0.1.22: unchanged; Scan Processor keeps its reviewed formats (its phases and per-line yokes in 32-bit float). 0.1.23: unchanged; Voxel Space keeps its reviewed formats (its camera and rows in 32-bit float, its land and eased colors in 16-bit). 0.1.24: unchanged; Temporal Resonator keeps its reviewed formats (its strings in 16-bit float, its stamps and taps in 32-bit). 0.1.25: unchanged; Caustics keeps its reviewed formats (its water and surface in 32-bit float, its light in 16-bit). 0.1.27: unchanged; Relight keeps its reviewed formats (its lights and surface in 32-bit float, its light in 16-bit). + view and label are 8-bit (the pictures); the Null TOPs pass each effect's own format; no data textures of its own (the effects inside keep their reviewed formats). 0.1.3: unchanged; Fluid keeps its reviewed formats. 0.1.4: unchanged; VHS + CRT keeps its reviewed formats. 0.1.5: unchanged; VHS + CRT keeps its reviewed formats. 0.1.7: unchanged; Bent Cam keeps its reviewed formats. 0.1.8: unchanged. 0.1.9: unchanged; Slit-Scan keeps its reviewed formats. 0.1.10: unchanged; Pixel Sort keeps its reviewed formats. 0.1.11: unchanged; Paint keeps its reviewed formats. 0.1.13: unchanged; Reaction-Diffusion keeps its reviewed formats (its pattern in 32-bit float). 0.1.14: unchanged; Pencil keeps its reviewed formats (its strokes' offsets in 32-bit float). 0.1.15: unchanged; Motion Magnify keeps its reviewed formats (its displacement in 32-bit float). 0.1.16: unchanged; Video Vocoder keeps its reviewed formats (its pyramid and passes in 32-bit float). 0.1.17: unchanged; Detail keeps its reviewed formats (its pyramids and passes in 32-bit float). 0.1.18: unchanged; Chladni Sand keeps its reviewed formats (its plate and sand in 32-bit float). 0.1.19: unchanged; Pixel Art keeps its reviewed formats (its grid's data textures in 32-bit float). 0.1.20: unchanged; Lens Flare keeps its reviewed formats (its tiles, lights, ghost table and the starburst's transform in 32-bit float). 0.1.21: unchanged; Words keeps its reviewed formats (its word tables, cells, field and slots in 32-bit float). 0.1.22: unchanged; Scan Processor keeps its reviewed formats (its phases and per-line yokes in 32-bit float). 0.1.23: unchanged; Voxel Space keeps its reviewed formats (its camera and rows in 32-bit float, its land and eased colors in 16-bit). 0.1.24: unchanged; Temporal Resonator keeps its reviewed formats (its strings in 16-bit float, its stamps and taps in 32-bit). 0.1.25: unchanged; Caustics keeps its reviewed formats (its water and surface in 32-bit float, its light in 16-bit). 0.1.27: unchanged; Relight keeps its reviewed formats (its lights and surface in 32-bit float, its light in 16-bit). 0.1.30: unchanged. Every effect inside is rebuilt for lib's new parts (warmth(), incompressible(), spectrum.glsl); dev/probes/structure_diff.py finds only shader text changed (the headers of Fluid's divergence, pressure and velocity, and Lens Flare's burst with the CIE fit now from lib). - [x] C7 The performance note on the docs page was measured on the reference machine at 1280 x 720, not estimated. - dev/perf.py on the 0.1.0 build with ascii, blob_track, datamosh, glitch, particles and point_cloud inside (MacBook Pro M2 Pro 16 GB, macOS 15.6, TouchDesigner 2025.33230), 1280 x 720, the dancer with its depth on input 3, the demo paused: Cycle (default) GPU 3.74 ms (95th percentile 8.55, two effects during a fade), CPU 1.39 ms (2.03); 60 fps held. Grid is gone: with six effects it took 9.95 ms GPU and grew with every effect. Remeasured whenever the effects inside change. dev/perf.py on the 0.1.1 build, the demo paused: Cycle GPU 3.69 ms (95th percentile 9.13, two effects during a fade), CPU 1.35 ms (1.92); 60 fps held. dev/perf.py on the 0.1.2 build, three runs, the demo paused: Cycle GPU 4.35, 4.23 and 3.05 ms (95th percentile 7.37, 7.45, 5.88), CPU 1.12, 1.66, 1.13 ms, 59-60 fps; the docs page now gives the medians, 4.2 ms (7.4) and 1.1 ms (1.6). dev/perf.py on the 0.1.3 build (seven effects inside), three runs, the demo idle: Cycle GPU 4.10, 4.05 and 4.08 ms (95th percentile 8.02, 7.89, 7.10), CPU 1.53, 1.48, 1.19 ms, 59 fps; the docs page gives the medians, 4.1 ms (7.9) and 1.5 ms (1.7). One effect runs at a time, so adding Fluid didn't add its cost. dev/perf.py on the 0.1.4 build (eight effects inside), three runs: Cycle GPU 4.20, 3.96 and 3.99 ms (95th percentile 8.36, 6.95, 7.51), CPU 1.38, 1.19, 1.19 ms, 59 fps; the docs page gives the medians, 4.0 ms (7.5) and 1.2 ms (1.7). One effect runs at a time, so adding VHS + CRT didn't add its cost. dev/perf.py on the 0.1.5 build, three runs: Cycle GPU 3.52, 3.55 and 3.69 ms (95th percentile 6.26, 7.38, 7.70), CPU 1.20, 1.44, 1.36 ms, 59 fps; the docs page gives the medians, 3.5 ms (7.4) and 1.4 ms (1.7). VHS + CRT inside runs its CRT, at its defaults. dev/perf.py on the 0.1.7 build (nine effects inside), three runs: Cycle GPU 3.45, 3.65 and 3.37 ms (95th percentile 6.92, 7.12, 7.16), CPU 1.07, 1.14, 1.08 ms, 59 fps; the docs page gives the medians, 3.5 ms (7.1) and 1.1 ms (1.3). One effect runs at a time, so adding Bent Cam didn't add its cost. 0.1.8: not re-measured: Bent Cam 0.1.1 costs what 0.1.0 did (1.44 ms GPU, its review C7), and one effect runs at a time. 0.1.9 (ten effects), dev/perf.py three runs: Cycle GPU 4.45, 3.85, 3.52 ms (95th percentile 9.60, 7.46, 6.95, two effects during a fade), CPU 1.47, 1.12, 1.12 ms (1.83, 1.34, 1.35), 59 fps held each time; the page gives the medians, 3.9 ms (7.5) and 1.1 ms (1.3). Slit-Scan's history keeps about 546 MB of GPU memory at 1280 x 720 once it has shown (its page, Memory). 0.1.10 (eleven effects), dev/perf.py three runs: Cycle GPU 3.66, 4.07, 3.78 ms (95th percentile 7.61, 9.44, 8.24), CPU 1.22, 1.55, 1.13 ms (1.84, 1.96, 1.39), 59 fps held each time; the page gives the medians, 3.8 ms (8.2) and 1.2 ms (1.8). Pixel Sort adds 2.7 ms GPU while it shows (its page). 0.1.11 (twelve effects), dev/perf.py three runs with the demo paused: Cycle GPU 3.98, 4.01, 4.12 ms (95th percentile 7.08, 6.90, 6.99), CPU 1.08, 1.07, 1.07 ms (1.25, 1.25, 1.24), 59 fps held each time; the page gives the medians, 4.0 ms (7.0) and 1.1 ms (1.2). Paint adds about 5.3 ms GPU while it shows (its page). 0.1.12 embeds every effect as rebuilt with its network laid out (Network Boxes, positions), the unused compute shaders, and lays out its own network the same way; the README holds one line per paragraph. dev/probes/structure_diff.py against 0.1.11 finds every other operator, parameter, expression and wire unchanged, so these numbers stand. 0.1.13 (thirteen effects, every effect rebuilt for lib's tensor easing), dev/perf.py three runs with the demo paused: Cycle GPU 4.01, 4.30, 4.32 ms (95th percentile 7.91, 8.90, 9.25), CPU 1.10, 1.09, 1.09 ms (1.32, 1.29, 1.26), 59 fps held each time; the page gives the medians, 4.3 ms (8.9) and 1.1 ms (1.3). Reaction-Diffusion adds about 4.0 ms GPU while it shows (its page). 0.1.14 (fourteen effects, every effect rebuilt for lib's paper.glsl), dev/perf.py three runs with the demo paused: Cycle GPU 4.27, 4.18, 4.22 ms (95th percentile 7.20, 7.76, 7.76), CPU 1.12, 1.13, 1.13 ms (1.38, 1.44, 1.38), 59 fps held each time; the page gives the medians, 4.2 ms (7.8) and 1.1 ms (1.4). Pencil adds about 5.3 ms GPU while it shows (its page). 0.1.15 (fifteen effects), dev/perf.py three runs with the demo paused: Cycle GPU 4.57, 4.57, 4.56 ms (95th percentile 9.73, 9.96, 9.76), CPU 1.31, 1.27, 1.27 ms (1.56, 1.53, 1.50), 59 fps held each time; the page gives the medians, 4.6 ms (9.8) and 1.3 ms (1.5). Motion Magnify adds about 3.6 ms GPU while it shows (its page). 0.1.16 (sixteen effects, every effect rebuilt for lib's bands and pyramid), dev/perf.py three runs with the demo paused: Cycle GPU 4.74, 4.65, 4.73 ms (95th percentile 10.38, 10.05, 10.28), CPU 1.30, 1.29, 1.29 ms (1.54, 1.55, 1.54), 59 fps held each time; the page gives the medians, 4.7 ms (10.3) and 1.3 ms (1.5). Video Vocoder adds about 0.9 ms GPU while it shows (its page). 0.1.17 (seventeen effects, every effect rebuilt for lib pyramid()'s prefix), dev/perf.py three runs with the demo paused: Cycle GPU 5.32, 5.30, 5.38 ms (95th percentile 10.16, 10.06, 10.83), CPU 1.35, 1.33, 1.45 ms (1.35, 1.37, 1.49), 59 fps held each time; the page gives the medians, 5.3 ms (10.2) and 1.4 ms (1.4). The mean follows which effects show in the 10 s measured (Blob Track's blobs was the costliest operator in all three); Detail adds about 2.9 ms GPU while it shows (its page). 0.1.18 (eighteen effects), dev/perf.py three runs with the demo paused: Cycle GPU 3.94, 4.24, 4.54 ms (95th percentile 6.60, 7.75, 7.69), CPU 1.31, 1.49, 1.50 ms (1.36, 1.54, 1.64), 58-59 fps; the page gives the medians, 4.2 ms (7.7) and 1.5 ms (1.5). The mean follows which effects show in the 10 s measured; Chladni Sand adds about 0.55 ms GPU while it shows (its page). 0.1.19 (nineteen effects), dev/perf.py three runs with the demo paused: Cycle GPU 3.91, 3.96, 3.92 ms (95th percentile 7.43, 7.19, 6.97), CPU 1.31, 1.32, 1.33 ms (1.38, 1.41, 1.49), 59 fps; the page gives the medians, 3.9 ms (7.2) and 1.3 ms (1.4). The mean follows which effects show in the 10 s measured; Pixel Art adds about 2.3 ms GPU while it shows (its page). 0.1.20 (twenty effects), dev/perf.py three runs with the demo paused: Cycle GPU 4.58, 4.63, 4.66 ms (95th percentile 8.28, 8.25, 9.24), CPU 1.34, 1.35, 1.40 ms (1.39, 1.45, 1.48), 59 fps; the page gives the medians, 4.6 ms (8.3) and 1.4 ms (1.5). The mean follows which effects show in the 10 s measured (Blob Track's blobs was the costliest operator, 1.6-1.8 ms); Lens Flare adds about 1.4 ms GPU while it shows (its page). 0.1.21 (twenty-one effects), dev/perf.py three runs with the demo paused: Cycle GPU 4.24, 4.32, 4.30 ms (95th percentile 6.92, 6.98, 6.78), CPU 1.40, 1.37, 1.37 ms (1.49, 1.48, 1.47), 59 fps; the page gives the medians, 4.3 ms (6.9) and 1.4 ms (1.5). The mean follows which effects show in the 10 s measured; Words adds 2.4 ms GPU while it shows at its defaults (its page). 0.1.22 (twenty-two effects), dev/perf.py three runs with the demo paused: Cycle GPU 4.14, 4.20, 4.24 ms (95th percentile 7.42, 7.56, 7.62), CPU 1.44, 1.42, 1.40 ms (1.61, 1.52, 1.49), 59 fps; the page gives the medians, 4.2 ms (7.6) and 1.4 ms (1.5). The mean follows which effects show in the 10 s measured (Blob Track's blobs was the costliest operator, about 1.2 ms); Scan Processor adds about 0.9 ms GPU while it shows (its page). 0.1.23 (twenty-three effects), dev/perf.py three runs with the demo paused: Cycle GPU 4.08, 4.15, 4.04 ms (95th percentile 7.34, 7.27, 7.22), CPU 1.39, 1.38, 1.38 ms (1.45, 1.46, 1.43), 59 fps; the page gives the medians, 4.1 ms (7.3) and 1.4 ms (1.5). Blob Track's blobs is still the costliest operator (about 1.1 ms); Voxel Space adds about 2.2 ms GPU while it shows (its page). 0.1.24 (twenty-four effects), dev/perf.py three runs with the demo paused: Cycle GPU 4.53, 4.40, 4.37 ms (95th percentile 8.35, 7.28, 7.14), CPU 1.59, 1.44, 1.44 ms (1.74, 1.59, 1.62), 59 fps; the page gives the medians, 4.4 ms (7.3) and 1.4 ms (1.6). Blob Track's blobs is still the costliest operator (about 1.3 ms); Temporal Resonator adds about 1.6 ms GPU while it shows and keeps 161 MB of GPU memory once it has (its page, and the showreel's beside Slit-Scan's). 0.1.25 (twenty-five effects), dev/perf.py three runs with the demo paused: Cycle GPU 4.23, 4.21, 4.21 ms (95th percentile 9.24, 9.02, 8.99), CPU 1.37, 1.35, 1.35 ms (1.54, 1.51, 1.52), 59 fps; the page gives the medians, 4.2 ms (9.0) and 1.4 ms (1.5). Blob Track's blobs is still the costliest operator (about 1.3 ms); Caustics adds about 2.2 ms GPU while it shows (its page), and the 95th percentile rose with fades into and out of it. 0.1.26: not re-measured. The passes Caustics and Temporal Resonator left out never cooked (their reviews, C5). 0.1.27 (twenty-six effects), dev/perf.py three runs with the demo paused: Cycle GPU 4.07, 3.86, 3.88 ms (95th percentile 8.68, 8.32, 8.08), CPU 1.40, 1.28, 1.28 ms (1.57, 1.42, 1.43), 59 fps; the page gives the medians, 3.9 ms (8.3) and 1.3 ms (1.4). Blob Track's blobs is still the costliest operator (about 1.3 ms); Relight adds about 4.6 ms GPU while it shows (its page). 0.1.28 (twenty-seven effects), dev/perf.py three runs with the demo paused: Cycle GPU 3.89, 3.90, 3.92 ms (95th percentile 8.36, 8.38, 8.60), CPU 1.32, 1.33, 1.33 ms, 59 fps each; the page's 3.9 and 1.3 ms stand. 0.1.29: not re-measured. Particles shows Emit here, whose code path and cost are unchanged (its review, C7), and its new middle_in passes a black Constant TOP in Emit. + dev/perf.py on the 0.1.0 build with ascii, blob_track, datamosh, glitch, particles and point_cloud inside (MacBook Pro M2 Pro 16 GB, macOS 15.6, TouchDesigner 2025.33230), 1280 x 720, the dancer with its depth on input 3, the demo paused: Cycle (default) GPU 3.74 ms (95th percentile 8.55, two effects during a fade), CPU 1.39 ms (2.03); 60 fps held. Grid is gone: with six effects it took 9.95 ms GPU and grew with every effect. Remeasured whenever the effects inside change. dev/perf.py on the 0.1.1 build, the demo paused: Cycle GPU 3.69 ms (95th percentile 9.13, two effects during a fade), CPU 1.35 ms (1.92); 60 fps held. dev/perf.py on the 0.1.2 build, three runs, the demo paused: Cycle GPU 4.35, 4.23 and 3.05 ms (95th percentile 7.37, 7.45, 5.88), CPU 1.12, 1.66, 1.13 ms, 59-60 fps; the docs page now gives the medians, 4.2 ms (7.4) and 1.1 ms (1.6). dev/perf.py on the 0.1.3 build (seven effects inside), three runs, the demo idle: Cycle GPU 4.10, 4.05 and 4.08 ms (95th percentile 8.02, 7.89, 7.10), CPU 1.53, 1.48, 1.19 ms, 59 fps; the docs page gives the medians, 4.1 ms (7.9) and 1.5 ms (1.7). One effect runs at a time, so adding Fluid didn't add its cost. dev/perf.py on the 0.1.4 build (eight effects inside), three runs: Cycle GPU 4.20, 3.96 and 3.99 ms (95th percentile 8.36, 6.95, 7.51), CPU 1.38, 1.19, 1.19 ms, 59 fps; the docs page gives the medians, 4.0 ms (7.5) and 1.2 ms (1.7). One effect runs at a time, so adding VHS + CRT didn't add its cost. dev/perf.py on the 0.1.5 build, three runs: Cycle GPU 3.52, 3.55 and 3.69 ms (95th percentile 6.26, 7.38, 7.70), CPU 1.20, 1.44, 1.36 ms, 59 fps; the docs page gives the medians, 3.5 ms (7.4) and 1.4 ms (1.7). VHS + CRT inside runs its CRT, at its defaults. dev/perf.py on the 0.1.7 build (nine effects inside), three runs: Cycle GPU 3.45, 3.65 and 3.37 ms (95th percentile 6.92, 7.12, 7.16), CPU 1.07, 1.14, 1.08 ms, 59 fps; the docs page gives the medians, 3.5 ms (7.1) and 1.1 ms (1.3). One effect runs at a time, so adding Bent Cam didn't add its cost. 0.1.8: not re-measured: Bent Cam 0.1.1 costs what 0.1.0 did (1.44 ms GPU, its review C7), and one effect runs at a time. 0.1.9 (ten effects), dev/perf.py three runs: Cycle GPU 4.45, 3.85, 3.52 ms (95th percentile 9.60, 7.46, 6.95, two effects during a fade), CPU 1.47, 1.12, 1.12 ms (1.83, 1.34, 1.35), 59 fps held each time; the page gives the medians, 3.9 ms (7.5) and 1.1 ms (1.3). Slit-Scan's history keeps about 546 MB of GPU memory at 1280 x 720 once it has shown (its page, Memory). 0.1.10 (eleven effects), dev/perf.py three runs: Cycle GPU 3.66, 4.07, 3.78 ms (95th percentile 7.61, 9.44, 8.24), CPU 1.22, 1.55, 1.13 ms (1.84, 1.96, 1.39), 59 fps held each time; the page gives the medians, 3.8 ms (8.2) and 1.2 ms (1.8). Pixel Sort adds 2.7 ms GPU while it shows (its page). 0.1.11 (twelve effects), dev/perf.py three runs with the demo paused: Cycle GPU 3.98, 4.01, 4.12 ms (95th percentile 7.08, 6.90, 6.99), CPU 1.08, 1.07, 1.07 ms (1.25, 1.25, 1.24), 59 fps held each time; the page gives the medians, 4.0 ms (7.0) and 1.1 ms (1.2). Paint adds about 5.3 ms GPU while it shows (its page). 0.1.12 embeds every effect as rebuilt with its network laid out (Network Boxes, positions), the unused compute shaders, and lays out its own network the same way; the README holds one line per paragraph. dev/probes/structure_diff.py against 0.1.11 finds every other operator, parameter, expression and wire unchanged, so these numbers stand. 0.1.13 (thirteen effects, every effect rebuilt for lib's tensor easing), dev/perf.py three runs with the demo paused: Cycle GPU 4.01, 4.30, 4.32 ms (95th percentile 7.91, 8.90, 9.25), CPU 1.10, 1.09, 1.09 ms (1.32, 1.29, 1.26), 59 fps held each time; the page gives the medians, 4.3 ms (8.9) and 1.1 ms (1.3). Reaction-Diffusion adds about 4.0 ms GPU while it shows (its page). 0.1.14 (fourteen effects, every effect rebuilt for lib's paper.glsl), dev/perf.py three runs with the demo paused: Cycle GPU 4.27, 4.18, 4.22 ms (95th percentile 7.20, 7.76, 7.76), CPU 1.12, 1.13, 1.13 ms (1.38, 1.44, 1.38), 59 fps held each time; the page gives the medians, 4.2 ms (7.8) and 1.1 ms (1.4). Pencil adds about 5.3 ms GPU while it shows (its page). 0.1.15 (fifteen effects), dev/perf.py three runs with the demo paused: Cycle GPU 4.57, 4.57, 4.56 ms (95th percentile 9.73, 9.96, 9.76), CPU 1.31, 1.27, 1.27 ms (1.56, 1.53, 1.50), 59 fps held each time; the page gives the medians, 4.6 ms (9.8) and 1.3 ms (1.5). Motion Magnify adds about 3.6 ms GPU while it shows (its page). 0.1.16 (sixteen effects, every effect rebuilt for lib's bands and pyramid), dev/perf.py three runs with the demo paused: Cycle GPU 4.74, 4.65, 4.73 ms (95th percentile 10.38, 10.05, 10.28), CPU 1.30, 1.29, 1.29 ms (1.54, 1.55, 1.54), 59 fps held each time; the page gives the medians, 4.7 ms (10.3) and 1.3 ms (1.5). Video Vocoder adds about 0.9 ms GPU while it shows (its page). 0.1.17 (seventeen effects, every effect rebuilt for lib pyramid()'s prefix), dev/perf.py three runs with the demo paused: Cycle GPU 5.32, 5.30, 5.38 ms (95th percentile 10.16, 10.06, 10.83), CPU 1.35, 1.33, 1.45 ms (1.35, 1.37, 1.49), 59 fps held each time; the page gives the medians, 5.3 ms (10.2) and 1.4 ms (1.4). The mean follows which effects show in the 10 s measured (Blob Track's blobs was the costliest operator in all three); Detail adds about 2.9 ms GPU while it shows (its page). 0.1.18 (eighteen effects), dev/perf.py three runs with the demo paused: Cycle GPU 3.94, 4.24, 4.54 ms (95th percentile 6.60, 7.75, 7.69), CPU 1.31, 1.49, 1.50 ms (1.36, 1.54, 1.64), 58-59 fps; the page gives the medians, 4.2 ms (7.7) and 1.5 ms (1.5). The mean follows which effects show in the 10 s measured; Chladni Sand adds about 0.55 ms GPU while it shows (its page). 0.1.19 (nineteen effects), dev/perf.py three runs with the demo paused: Cycle GPU 3.91, 3.96, 3.92 ms (95th percentile 7.43, 7.19, 6.97), CPU 1.31, 1.32, 1.33 ms (1.38, 1.41, 1.49), 59 fps; the page gives the medians, 3.9 ms (7.2) and 1.3 ms (1.4). The mean follows which effects show in the 10 s measured; Pixel Art adds about 2.3 ms GPU while it shows (its page). 0.1.20 (twenty effects), dev/perf.py three runs with the demo paused: Cycle GPU 4.58, 4.63, 4.66 ms (95th percentile 8.28, 8.25, 9.24), CPU 1.34, 1.35, 1.40 ms (1.39, 1.45, 1.48), 59 fps; the page gives the medians, 4.6 ms (8.3) and 1.4 ms (1.5). The mean follows which effects show in the 10 s measured (Blob Track's blobs was the costliest operator, 1.6-1.8 ms); Lens Flare adds about 1.4 ms GPU while it shows (its page). 0.1.21 (twenty-one effects), dev/perf.py three runs with the demo paused: Cycle GPU 4.24, 4.32, 4.30 ms (95th percentile 6.92, 6.98, 6.78), CPU 1.40, 1.37, 1.37 ms (1.49, 1.48, 1.47), 59 fps; the page gives the medians, 4.3 ms (6.9) and 1.4 ms (1.5). The mean follows which effects show in the 10 s measured; Words adds 2.4 ms GPU while it shows at its defaults (its page). 0.1.22 (twenty-two effects), dev/perf.py three runs with the demo paused: Cycle GPU 4.14, 4.20, 4.24 ms (95th percentile 7.42, 7.56, 7.62), CPU 1.44, 1.42, 1.40 ms (1.61, 1.52, 1.49), 59 fps; the page gives the medians, 4.2 ms (7.6) and 1.4 ms (1.5). The mean follows which effects show in the 10 s measured (Blob Track's blobs was the costliest operator, about 1.2 ms); Scan Processor adds about 0.9 ms GPU while it shows (its page). 0.1.23 (twenty-three effects), dev/perf.py three runs with the demo paused: Cycle GPU 4.08, 4.15, 4.04 ms (95th percentile 7.34, 7.27, 7.22), CPU 1.39, 1.38, 1.38 ms (1.45, 1.46, 1.43), 59 fps; the page gives the medians, 4.1 ms (7.3) and 1.4 ms (1.5). Blob Track's blobs is still the costliest operator (about 1.1 ms); Voxel Space adds about 2.2 ms GPU while it shows (its page). 0.1.24 (twenty-four effects), dev/perf.py three runs with the demo paused: Cycle GPU 4.53, 4.40, 4.37 ms (95th percentile 8.35, 7.28, 7.14), CPU 1.59, 1.44, 1.44 ms (1.74, 1.59, 1.62), 59 fps; the page gives the medians, 4.4 ms (7.3) and 1.4 ms (1.6). Blob Track's blobs is still the costliest operator (about 1.3 ms); Temporal Resonator adds about 1.6 ms GPU while it shows and keeps 161 MB of GPU memory once it has (its page, and the showreel's beside Slit-Scan's). 0.1.25 (twenty-five effects), dev/perf.py three runs with the demo paused: Cycle GPU 4.23, 4.21, 4.21 ms (95th percentile 9.24, 9.02, 8.99), CPU 1.37, 1.35, 1.35 ms (1.54, 1.51, 1.52), 59 fps; the page gives the medians, 4.2 ms (9.0) and 1.4 ms (1.5). Blob Track's blobs is still the costliest operator (about 1.3 ms); Caustics adds about 2.2 ms GPU while it shows (its page), and the 95th percentile rose with fades into and out of it. 0.1.26: not re-measured. The passes Caustics and Temporal Resonator left out never cooked (their reviews, C5). 0.1.27 (twenty-six effects), dev/perf.py three runs with the demo paused: Cycle GPU 4.07, 3.86, 3.88 ms (95th percentile 8.68, 8.32, 8.08), CPU 1.40, 1.28, 1.28 ms (1.57, 1.42, 1.43), 59 fps; the page gives the medians, 3.9 ms (8.3) and 1.3 ms (1.4). Blob Track's blobs is still the costliest operator (about 1.3 ms); Relight adds about 4.6 ms GPU while it shows (its page). 0.1.28 (twenty-seven effects), dev/perf.py three runs with the demo paused: Cycle GPU 3.89, 3.90, 3.92 ms (95th percentile 8.36, 8.38, 8.60), CPU 1.32, 1.33, 1.33 ms, 59 fps each; the page's 3.9 and 1.3 ms stand. 0.1.29: not re-measured. Particles shows Emit here, whose code path and cost are unchanged (its review, C7), and its new middle_in passes a black Constant TOP in Emit. 0.1.30: not re-measured. The effects inside run the same code: only shader comments changed (Fluid's projection, Lens Flare's starburst). - [x] D5 The README DAT and docs page say how to update: dragging a `.tox` in embeds a copy, so updating means dragging in the new one (or pointing External .tox at it). [A word about external files](https://derivative.ca/community-post/word-about-external-files-202310k-builds/68166) - The README DAT's UPDATING section and the docs page's Use it section say to delete the copy and drag in the new showreel.tox; both say a new effect in the library means a new showreel. 0.1.3: unchanged. 0.1.4: unchanged. 0.1.5: unchanged. 0.1.7: unchanged. 0.1.8: unchanged. 0.1.9: unchanged. 0.1.10: unchanged. 0.1.11: unchanged. 0.1.12: unchanged. 0.1.13: unchanged. 0.1.14: unchanged. 0.1.15: unchanged. 0.1.16: unchanged. 0.1.17: unchanged. 0.1.18: unchanged. 0.1.19: unchanged. 0.1.20: unchanged. 0.1.21: unchanged. 0.1.22: unchanged. 0.1.23: unchanged. 0.1.24: unchanged. 0.1.25: unchanged. 0.1.26: unchanged. 0.1.27: unchanged. 0.1.29: unchanged. + The README DAT's UPDATING section and the docs page's Use it section say to delete the copy and drag in the new showreel.tox; both say a new effect in the library means a new showreel. 0.1.3: unchanged. 0.1.4: unchanged. 0.1.5: unchanged. 0.1.7: unchanged. 0.1.8: unchanged. 0.1.9: unchanged. 0.1.10: unchanged. 0.1.11: unchanged. 0.1.12: unchanged. 0.1.13: unchanged. 0.1.14: unchanged. 0.1.15: unchanged. 0.1.16: unchanged. 0.1.17: unchanged. 0.1.18: unchanged. 0.1.19: unchanged. 0.1.20: unchanged. 0.1.21: unchanged. 0.1.22: unchanged. 0.1.23: unchanged. 0.1.24: unchanged. 0.1.25: unchanged. 0.1.26: unchanged. 0.1.27: unchanged. 0.1.29: unchanged. 0.1.30: unchanged. Every effect inside is rebuilt for lib's new parts (warmth(), incompressible(), spectrum.glsl); dev/probes/structure_diff.py finds only shader text changed (the headers of Fluid's divergence, pressure and velocity, and Lens Flare's burst with the CIE fit now from lib). - [x] D6 Provenance: the README DAT's CREDITS and the docs page's Credits and license name every paper, technique, project, code, font, model and media the effect uses, follows or was built after, with links and licenses, and call original only what was designed here and has no earlier source after a search. Code recalled from elsewhere that can't be traced is replaced, never called original; code taken from elsewhere is also credited where it sits (a comment in the shader or script). D3 and check_tox.py only see that the sections exist. - CREDITS and the docs page credit the effects inside to their own pages, the built-in dancer clip (Kevin Raymond, CC BY 4.0, AI-generated with Krea 2 and Wan 2.2) and her depth (Depth Anything 3 Base, Apache 2.0), the 5 x 7 font and WCAG 2.3.1. 0.1.1: the beat clock (Cycle's bars on raw audio) is credited in both as inter-onset-interval tempo induction, the idea behind Dixon 2001, followed, not copied. 0.1.2: her depth is now Depth Anything 3's Mono-Large model (Apache 2.0), one frame at a time, held steady from frame to frame by a scale fit written for this library; credited so in both. 0.1.3: Fluid is credited on its own page like the others; the docs page's preview (the README's hero) now includes Fluid's take, recorded like the rest (b1_dancer, m1x_ease_from_drop, 8 s). 0.1.4: VHS + CRT is credited on its own page like the others; the docs page's preview (the README's hero) now includes its take, recorded like the rest (b1_dancer, m1x_ease_from_drop, 8 s). 0.1.5: VHS + CRT 0.2.0 credits its screens on its own page like the rest; the preview and hero are unchanged (VHS + CRT at its defaults is the same CRT). 0.1.7: Bent Cam is credited on its own page like the others; the docs page's preview (the README's hero) now includes its take, recorded like the rest (b1_dancer, 8 s), in rows of three. 0.1.8: the preview (the README's hero) takes Bent Cam 0.1.1's new take, recorded like the rest. 0.1.9: Slit-Scan's credits are on its page and in its README; nothing new of the showreel's own. 0.1.10: Pixel Sort's credits are on its page and in its README; the docs page's preview (the README's hero) takes its take, recorded like the rest (b1_dancer, m1x_ease_from_drop, 8 s), in rows of 2-3-3-3-2. 0.1.11: Paint's credits are on its page and in its README; the docs page's preview (the README's hero) takes its take, recorded like the rest (b1_dancer, m1x_ease_from_drop, 8 s), in rows of 2-3-3-3-3. 0.1.13: Reaction-Diffusion's credits are on its page and in its README; the docs page's preview (the README's hero) takes its take, recorded like the rest (b1_dancer, m1x_ease_from_drop, 8 s) after letting it grow in for 5 s (dev/record.py settle=5), in rows of 2-3-2-3-2-3. 0.1.14: Pencil's credits are on its page and in its README; the hero takes its take, recorded like the rest (b1_dancer, the song, 8 s), in rows of 2-3-3-3-3-2, at WebP quality 65 (7.2 MB; Pencil's fine strokes compress poorly). 0.1.15: Motion Magnify's credits are on its page and in its README; the hero takes its take, recorded like the rest (b1_dancer, m1x_ease_from_drop, 8 s, with the input take again), in rows of 2-3-3-3-3-3, at WebP quality 65 (7.3 MB). 0.1.16: Video Vocoder's credits are on its page and in its README; the hero takes its take, recorded like the rest (b1_dancer, m1x_ease_from_drop, 8 s, with the input take again, after 5 s for its usual levels to form) but with its four fine octaves at 2, Drive 1 and Flashing Up to 3 per second: at its defaults the dancer, who has little fine texture, looked like the input at the hero's size (detail 0.6 to 1.4 times the input's; now 1.6 to 2.6), in rows of 2-3-3-3-3-2-2 (the clip and its depth map, the inputs, alone on top), at WebP quality 65 (7.7 MB). 0.1.17: Detail's credits are on its page and in its README; the hero takes its take, recorded like the rest (b1_dancer, m1x_ease_from_drop, 8 s, after 5 s for its usual loudness to form) but with Detail 2.5 and Edge Contrast 0.35: at its defaults the dancer's haze changed too little to read at the hero's size. The hero is now 2-3-3-3-3-3-2 (the inputs alone on top); the social preview is unchanged ('& MORE'). 0.1.18: Chladni Sand's credits are on its page and in its README; the hero takes its take, recorded like the rest (b1_dancer, m1x_ease_from_drop, 8 s, after 5 s) but with Flashing at Up to 3 per second: at Calm its figure showed on too few strikes to read in the 6 s preview, and Settle 0.8 as well hid the dancer. The hero is now 2 then six rows of three (the inputs alone on top); the social preview is unchanged ('& MORE'). 0.1.19: Pixel Art's credits are on its page and in its README; the hero takes its take at its defaults, recorded like the rest (b1_dancer, m1x_ease_from_drop, 8 s). The hero is now 2, then five rows of three and two of two (the inputs alone on top), its WebP at quality 62 to stay near its old size (8.1 MB); the social preview is unchanged ('& MORE'). 0.1.20: Lens Flare's credits are on its page and in its README; the hero takes its take at its defaults, recorded like the rest (b1_dancer, m1x_ease_from_drop, 8 s). The hero is now 2, then six rows of three and one of two (the inputs alone on top), its WebP at quality 62 (8.3 MB); the social preview is unchanged ('& MORE'). 0.1.21: Words' credits are on its page and in its README; the hero takes its take at its defaults (Cloud), recorded like the rest (b1_dancer, m1x_ease_from_drop, 8 s). The hero is now 2, then seven rows of three (the inputs alone on top), its WebP at quality 62 (8.3 MB); the social preview is unchanged ('& MORE'). After 0.1.21, the hero's Reaction-Diffusion take was recorded again, like the rest (b1_dancer, m1x_ease_from_drop, 8 s, after 5 s to grow in), but with Size at Fine, Grow On at Depth and Colors at Light on Dark: at its defaults the stripes filled the haze and the dark warehouse alike, and the dancer was lost in them. Now she alone is drawn, white stripes curving along her body on black. The hero's WebP is still at quality 62 (7.0 MB); the social preview is unchanged ('& MORE'). 0.1.22: Scan Processor's credits are on its page and in its README; the hero takes its take at its defaults (Wobbulator), recorded like the rest (b1_dancer, m1x_ease_from_drop, 8 s). The hero is now 2, then six rows of three and two of two (the inputs alone on top), its WebP at quality 85 (15.1 MB): at 62 the dark haze broke into blocks (Datamosh's tile worst), and 85 was as clean as 95 (31.8 MB). The social preview is unchanged ('& MORE'). 0.1.23: Voxel Space's credits are on its page and in its README; the hero takes its take at its defaults (Fly), recorded like the rest (b1_dancer, m1x_ease_from_drop, 8 s). The hero is now 2, then seven rows of three and one of two, its WebP at quality 100 (46 MB): at 85 and at 95 the Voxel Space tile's gray sky, whose color drifts with the picture's, came out as a patchwork of near-identical blocks on some frames (lossy animated WebP keeps blocks that changed only a little from the frame before), while the take and the MP4 were clean; 98 saved little. The social preview is unchanged ('& MORE'). 0.1.24: Temporal Resonator's credits are on its page and in its README; the hero takes its take at its defaults, recorded like the rest (b1_dancer, m1x_ease_from_drop, 8 s). The hero is now 2, then eight rows of three, its WebP at quality 100 (48 MB), as 0.1.23's. 0.1.25: Caustics' credits are on its page and in its README; the hero takes its take at its defaults, recorded like the rest (b1_dancer, m1x_ease_from_drop, 8 s). The hero is now 2, then seven rows of three and two of two, its WebP at quality 100 (50 MB). 0.1.27: Relight's credits are on its page and in its README; the hero takes its take at its defaults, recorded like the rest (b1_dancer, m1x_ease_from_drop, 8 s). The hero is now 2, then eight rows of three and one of two, its WebP at quality 100 (52 MB). 0.1.28: Film Stock's credits are on its page and in its README; the hero takes its take recorded like the rest (b1_dancer, m1x_ease_from_drop, 8 s) but with Gauge at Super 8: at its default 16 mm the grain was too fine to read at the hero's size, and every other take was recorded again the same way, with the earlier special settings kept for Video Vocoder, Detail, Chladni Sand and Reaction-Diffusion. The hero is now 2, then nine rows of three, its WebP at quality 100. + CREDITS and the docs page credit the effects inside to their own pages, the built-in dancer clip (Kevin Raymond, CC BY 4.0, AI-generated with Krea 2 and Wan 2.2) and her depth (Depth Anything 3 Base, Apache 2.0), the 5 x 7 font and WCAG 2.3.1. 0.1.1: the beat clock (Cycle's bars on raw audio) is credited in both as inter-onset-interval tempo induction, the idea behind Dixon 2001, followed, not copied. 0.1.2: her depth is now Depth Anything 3's Mono-Large model (Apache 2.0), one frame at a time, held steady from frame to frame by a scale fit written for this library; credited so in both. 0.1.3: Fluid is credited on its own page like the others; the docs page's preview (the README's hero) now includes Fluid's take, recorded like the rest (b1_dancer, m1x_ease_from_drop, 8 s). 0.1.4: VHS + CRT is credited on its own page like the others; the docs page's preview (the README's hero) now includes its take, recorded like the rest (b1_dancer, m1x_ease_from_drop, 8 s). 0.1.5: VHS + CRT 0.2.0 credits its screens on its own page like the rest; the preview and hero are unchanged (VHS + CRT at its defaults is the same CRT). 0.1.7: Bent Cam is credited on its own page like the others; the docs page's preview (the README's hero) now includes its take, recorded like the rest (b1_dancer, 8 s), in rows of three. 0.1.8: the preview (the README's hero) takes Bent Cam 0.1.1's new take, recorded like the rest. 0.1.9: Slit-Scan's credits are on its page and in its README; nothing new of the showreel's own. 0.1.10: Pixel Sort's credits are on its page and in its README; the docs page's preview (the README's hero) takes its take, recorded like the rest (b1_dancer, m1x_ease_from_drop, 8 s), in rows of 2-3-3-3-2. 0.1.11: Paint's credits are on its page and in its README; the docs page's preview (the README's hero) takes its take, recorded like the rest (b1_dancer, m1x_ease_from_drop, 8 s), in rows of 2-3-3-3-3. 0.1.13: Reaction-Diffusion's credits are on its page and in its README; the docs page's preview (the README's hero) takes its take, recorded like the rest (b1_dancer, m1x_ease_from_drop, 8 s) after letting it grow in for 5 s (dev/record.py settle=5), in rows of 2-3-2-3-2-3. 0.1.14: Pencil's credits are on its page and in its README; the hero takes its take, recorded like the rest (b1_dancer, the song, 8 s), in rows of 2-3-3-3-3-2, at WebP quality 65 (7.2 MB; Pencil's fine strokes compress poorly). 0.1.15: Motion Magnify's credits are on its page and in its README; the hero takes its take, recorded like the rest (b1_dancer, m1x_ease_from_drop, 8 s, with the input take again), in rows of 2-3-3-3-3-3, at WebP quality 65 (7.3 MB). 0.1.16: Video Vocoder's credits are on its page and in its README; the hero takes its take, recorded like the rest (b1_dancer, m1x_ease_from_drop, 8 s, with the input take again, after 5 s for its usual levels to form) but with its four fine octaves at 2, Drive 1 and Flashing Up to 3 per second: at its defaults the dancer, who has little fine texture, looked like the input at the hero's size (detail 0.6 to 1.4 times the input's; now 1.6 to 2.6), in rows of 2-3-3-3-3-2-2 (the clip and its depth map, the inputs, alone on top), at WebP quality 65 (7.7 MB). 0.1.17: Detail's credits are on its page and in its README; the hero takes its take, recorded like the rest (b1_dancer, m1x_ease_from_drop, 8 s, after 5 s for its usual loudness to form) but with Detail 2.5 and Edge Contrast 0.35: at its defaults the dancer's haze changed too little to read at the hero's size. The hero is now 2-3-3-3-3-3-2 (the inputs alone on top); the social preview is unchanged ('& MORE'). 0.1.18: Chladni Sand's credits are on its page and in its README; the hero takes its take, recorded like the rest (b1_dancer, m1x_ease_from_drop, 8 s, after 5 s) but with Flashing at Up to 3 per second: at Calm its figure showed on too few strikes to read in the 6 s preview, and Settle 0.8 as well hid the dancer. The hero is now 2 then six rows of three (the inputs alone on top); the social preview is unchanged ('& MORE'). 0.1.19: Pixel Art's credits are on its page and in its README; the hero takes its take at its defaults, recorded like the rest (b1_dancer, m1x_ease_from_drop, 8 s). The hero is now 2, then five rows of three and two of two (the inputs alone on top), its WebP at quality 62 to stay near its old size (8.1 MB); the social preview is unchanged ('& MORE'). 0.1.20: Lens Flare's credits are on its page and in its README; the hero takes its take at its defaults, recorded like the rest (b1_dancer, m1x_ease_from_drop, 8 s). The hero is now 2, then six rows of three and one of two (the inputs alone on top), its WebP at quality 62 (8.3 MB); the social preview is unchanged ('& MORE'). 0.1.21: Words' credits are on its page and in its README; the hero takes its take at its defaults (Cloud), recorded like the rest (b1_dancer, m1x_ease_from_drop, 8 s). The hero is now 2, then seven rows of three (the inputs alone on top), its WebP at quality 62 (8.3 MB); the social preview is unchanged ('& MORE'). After 0.1.21, the hero's Reaction-Diffusion take was recorded again, like the rest (b1_dancer, m1x_ease_from_drop, 8 s, after 5 s to grow in), but with Size at Fine, Grow On at Depth and Colors at Light on Dark: at its defaults the stripes filled the haze and the dark warehouse alike, and the dancer was lost in them. Now she alone is drawn, white stripes curving along her body on black. The hero's WebP is still at quality 62 (7.0 MB); the social preview is unchanged ('& MORE'). 0.1.22: Scan Processor's credits are on its page and in its README; the hero takes its take at its defaults (Wobbulator), recorded like the rest (b1_dancer, m1x_ease_from_drop, 8 s). The hero is now 2, then six rows of three and two of two (the inputs alone on top), its WebP at quality 85 (15.1 MB): at 62 the dark haze broke into blocks (Datamosh's tile worst), and 85 was as clean as 95 (31.8 MB). The social preview is unchanged ('& MORE'). 0.1.23: Voxel Space's credits are on its page and in its README; the hero takes its take at its defaults (Fly), recorded like the rest (b1_dancer, m1x_ease_from_drop, 8 s). The hero is now 2, then seven rows of three and one of two, its WebP at quality 100 (46 MB): at 85 and at 95 the Voxel Space tile's gray sky, whose color drifts with the picture's, came out as a patchwork of near-identical blocks on some frames (lossy animated WebP keeps blocks that changed only a little from the frame before), while the take and the MP4 were clean; 98 saved little. The social preview is unchanged ('& MORE'). 0.1.24: Temporal Resonator's credits are on its page and in its README; the hero takes its take at its defaults, recorded like the rest (b1_dancer, m1x_ease_from_drop, 8 s). The hero is now 2, then eight rows of three, its WebP at quality 100 (48 MB), as 0.1.23's. 0.1.25: Caustics' credits are on its page and in its README; the hero takes its take at its defaults, recorded like the rest (b1_dancer, m1x_ease_from_drop, 8 s). The hero is now 2, then seven rows of three and two of two, its WebP at quality 100 (50 MB). 0.1.27: Relight's credits are on its page and in its README; the hero takes its take at its defaults, recorded like the rest (b1_dancer, m1x_ease_from_drop, 8 s). The hero is now 2, then eight rows of three and one of two, its WebP at quality 100 (52 MB). 0.1.28: Film Stock's credits are on its page and in its README; the hero takes its take recorded like the rest (b1_dancer, m1x_ease_from_drop, 8 s) but with Gauge at Super 8: at its default 16 mm the grain was too fine to read at the hero's size, and every other take was recorded again the same way, with the earlier special settings kept for Video Vocoder, Detail, Chladni Sand and Reaction-Diffusion. The hero is now 2, then nine rows of three, its WebP at quality 100. 0.1.30: Fluid's projection shaders and Lens Flare's spectrum fit now sit in lib (incompressible(), spectrum.glsl), credited as before. - [x] A2 The flash limits were measured, not assumed: the output recorded with a beat-heavy input at each Flashing setting, and the flashes in any one second counted. - dev/flash.py + dev/flashcount.py on the 0.1.0 build with the six effects inside (particles and point_cloud getting the dancer's depth, Depth Anything 3 Base, on input 3, showing the clip's frame), one run at a time, stress beat of 8 beats, 2 bars and 1 drop a second (Cycle then moves every 2 s, the fastest it allows). Worst second, area rule: Cycle on the dancer Off 1.0, Calm 2.0, Up to 3 per second 3.0 (clip alone 1.0); on the plaza 0.5, 1.5, 3.0 (clip 0.5); at Every 1 on the dancer 1.0, 2.0, 2.5. Cycle's cut, before the fade, reached 3.5 at Up to 3 per second with six effects (FAIL); the 1 s fade fixed it. No red flashes. All within flashcount's bars. 0.1.1 (Datamosh's bloom and Particles' sparks gated, both new): Cycle on the dancer Off 1.0, Calm 2.0, Up to 3 per second 2.5 (clip 1.0); plaza 0.5, 1.5, 3.0 (clip 0.5); Every 1 on the dancer 1.0, 2.0, 2.5. No red flashes; each within its setting's rule. 0.1.2 (the new dancer depth, Particles 0.3.1, Point Cloud 0.1.2): Cycle on the dancer Off 1.0, Calm 2.0, Up to 3 per second 3.0 (clip 1.0), within WCAG's 3 as 0.1.0 was; the plaza carries no depth, so its numbers stand. 0.1.3 (Fluid added): Cycle on the dancer Off 1.0, Calm 2.0, Up to 3 per second 3.0 (clip 1.0), red 0.5 at Up to 3 per second (Glitch's color split, which measures 1.0 red on its own, happened to show in the 10 s; under WCAG's 3); plaza 0.5, 1.5, 2.5 (clip 0.5); Every 1 on the dancer 1.0, 2.0, 2.5. Each within its setting's rule. 0.1.4 (VHS + CRT added): Cycle on the dancer Off 1.0, Calm 2.0, Up to 3 per second 2.0 (clip 1.0); plaza 0.5, 1.5, 1.5 (clip 0.5); Every 1 on the dancer 1.0, 2.0, 2.5; red 0.5 at Up to 3 per second on the dancer and the plaza (Glitch's split), none otherwise. Each within its setting's rule. 0.1.5 (VHS + CRT 0.2.0, every effect rebuilt for the lib change): Cycle on the dancer Off 1.5, Calm 2.0, Up to 3 per second 2.5 (clip 1.0; Off within flashcount's 0.5 of measuring noise, which effects showed in the 10 s differs from 0.1.4's 1.0); plaza 0.5, 1.5, 1.0 (clip 0.5); Every 1 on the dancer 1.0, 2.0, 2.5; red 0.5 at Up to 3 per second on all three (Glitch's split), none otherwise. Each within its setting's rule. 0.1.6: not re-measured: only the shader clock changed, wrapped every 16384 s (board #4681), which changes its value, not what the effect does with it, apart from one jump each 4.6 h. 0.1.7 (Bent Cam added, every effect rebuilt for the lib change): Cycle on the dancer Off 1.0, Calm 1.5, Up to 3 per second 2.5 (clip 1.0); plaza 1.0, 1.5, 2.0 (clip 0.5; Off within flashcount's 0.5 of measuring noise); Every 1 on the dancer 1.0, 1.5, 2.5; red 1.0 at Up to 3 per second on the dancer and 0.5 on the plaza (Glitch's split, the only effect with red flashes on the dancer: Glitch alone measures 1.0 at Calm and Up to 3), none otherwise. Each within its setting's rule. 0.1.8 (Bent Cam 0.1.1 inside): Cycle on the dancer Off 1.0, Calm 1.5, Up to 3 per second 2.5 (clip 1.0); red 1.0 at Up to 3 per second (Glitch's split), none otherwise. Each within its setting's rule. The plaza and Every 1 not re-measured: only Bent Cam's beats changed, measured on its own page. 0.1.9 (Slit-Scan added): before the fix, Cycle on the dancer measured Up to 3 per second 3.5 (FAIL): a fade from Bent Cam into Blob Track with Blob Track flashing as fast as Up to 3 allows in the same second, a pairing the 10 s window hadn't caught before. Now Up to 3 per second holds the effects at Calm while Cycle fades and for a second after. After: dancer Off 1.0, Calm 1.5, Up to 3 per second 2.0 (clip 1.0); plaza 1.0, 1.0, 2.0 (clip 0.5; Off within flashcount's 0.5 of measuring noise); Every 1 on the dancer 1.0, 1.5, 2.0. No red flashes. Each within its setting's rule. 0.1.10 (eleven effects), on the final build: dancer Off 1.0, Calm 1.5, Up to 3 per second 2.5 (clip 1.0); plaza 1.0, 1.0, 1.5 (clip 0.5; Off within flashcount's 0.5 of measuring noise); Every 1 on the dancer 1.0, 1.5, 2.0. No red flashes. Each within its setting's rule. 0.1.11 (twelve effects), on the final build with the demo paused: dancer Off 1.0, Calm 1.5, Up to 3 per second 2.0 (clip 1.0); plaza 0.5, 1.0, 1.5 (clip 0.5); Every 1 on the dancer 1.0, 1.0, 2.0. No red flashes. Each within its setting's rule. 0.1.12 embeds every effect as rebuilt with its network laid out (Network Boxes, positions), the unused compute shaders, and lays out its own network the same way; the README holds one line per paragraph. dev/probes/structure_diff.py against 0.1.11 finds every other operator, parameter, expression and wire unchanged, so these numbers stand. 0.1.13 (thirteen effects), on the final build with the demo paused: dancer Off 1.0, Calm 1.5, Up to 3 per second 2.0 (clip 1.0); plaza 1.0, 1.0, 2.0 (clip 0.5; Off within flashcount's 0.5 of measuring noise); Every 1 on the dancer 1.0, 1.5, 2.0. No red flashes. Each within its setting's rule. 0.1.14 (fourteen effects), on the final build with the demo paused: dancer Off 1.0, Calm 1.5, Up to 3 per second 1.5 (clip 1.0); plaza 0.5, 1.0, 1.5 (clip 0.5); Every 1 on the dancer 1.0, 1.0, 1.5. No red flashes. Each within its setting's rule. 0.1.15 (fifteen effects), on the final build with the demo paused: dancer Off 1.0, Calm 1.5, Up to 3 per second 1.5 (clip 1.0); plaza 1.0, 1.0, 1.5 (clip 0.5; Off within flashcount's 0.5 of measuring noise); Every 1 on the dancer 1.0, 1.5, 1.5. No red flashes. Each within its setting's rule. 0.1.16 (sixteen effects), on the final build with the demo paused: dancer Off 1.0, Calm 1.5, Up to 3 per second 1.5 (clip 1.0); plaza 1.0, 0.5, 1.5 (clip 0.5; Off within flashcount's 0.5 of measuring noise); Every 1 on the dancer 1.0, 1.5, 1.5. No red flashes. Each within its setting's rule. Video Vocoder inside keeps its own Flashing (its review A2). 0.1.17 (seventeen effects), with the demo paused: dancer Off 1.0, Calm 1.0, Up to 3 per second 1.5 (clip 1.0); plaza 1.0, 0.5, 1.0 (clip 0.5; Off within flashcount's 0.5 of measuring noise); Every 1 on the dancer 1.0, 1.5, 1.0. No red flashes. Each within its setting's rule. Detail inside keeps its own Flashing (its review A2). 0.1.18 (eighteen effects), with the demo paused: dancer Off 1.0, Calm 2.0, Up to 3 per second 2.0 (clip 1.0; Calm at its full allowance); plaza 0.5, 1.0, 1.0 (clip 0.5); Every 1 on the dancer 1.0, 1.5, 1.5. No red flashes. Each within its setting's rule. Chladni Sand inside keeps its own Flashing (its review A2). 0.1.19 (nineteen effects), with the demo paused: dancer Off 1.0, Calm 1.5, Up to 3 per second 1.5 (clip 1.0); plaza 0.5, 1.0, 1.0 (clip 0.5); Every 1 on the dancer 1.0, 1.0, 1.5. No red flashes. Each within its setting's rule. Pixel Art inside keeps its own Flashing (its review A2). 0.1.20 (twenty effects), with the demo paused: dancer Off 1.0, Calm 2.0, Up to 3 per second 1.5 (clip 1.0); plaza 1.0, 1.0, 1.0 (clip 0.5); Every 1 on the dancer 1.0, 1.5, 1.5. No red flashes. Each within its setting's rule. Lens Flare inside keeps its own Flashing (its review A2). 0.1.21 (twenty-one effects), with the demo paused: dancer Off 1.0, Calm 2.0, Up to 3 per second 1.5 (clip 1.0); plaza 1.0, 1.0, 1.0 (clip 0.5); Every 1 on the dancer 1.0, 1.5, 1.5. No red flashes. Each within its setting's rule. Words inside keeps its own Flashing (its review A2). 0.1.22 (twenty-two effects), with the demo paused: dancer Off 1.0, Calm 2.0, Up to 3 per second 1.5 (clip 1.0); plaza 1.0, 1.0, 1.0 (clip 0.5); Every 1 on the dancer 1.0, 1.5 (red 0.5), 1.0. Each within its setting's rule (the plaza's Off at the half-flash margin). Scan Processor inside keeps its own Flashing (its review A2). 0.1.23 (twenty-three effects), with the demo paused: dancer Off 1.0, Calm 1.5, Up to 3 per second 1.5 (clip 1.0); plaza 1.0, 1.0, 1.0 (clip 0.5); Every 1 on the dancer 1.0, 1.0, 1.5. Each within its setting's rule (the plaza's Off at the half-flash margin). Voxel Space inside keeps its own Flashing (its review A2). 0.1.24 (twenty-four effects), with the demo paused: dancer Off 1.0, Calm 1.5, Up to 3 per second 1.5 (clip 1.0); plaza 1.0, 1.0, 1.0 (clip 0.5); Every 1 on the dancer 1.0, 1.5, 1.5. Each within its setting's rule (the plaza's Off at the half-flash margin). Temporal Resonator inside has no Flashing setting and adds nothing to its clips' own (its review A2). 0.1.25 (twenty-five effects), with the demo paused: dancer Off 1.0, Calm 2.0, Up to 3 per second 1.5 (clip 1.0); plaza 1.0, 1.0, 1.0 (clip 0.5); Every 1 on the dancer 1.0, 1.5, 1.0. Each within its setting's rule (the plaza's Off at the half-flash margin). Caustics inside has no Flashing setting and adds nothing to its clips' own beyond the margin (its review A2). 0.1.26: not re-measured. The passes Caustics and Temporal Resonator left out never cooked (their reviews, C5), so the output is the same. 0.1.27 (twenty-six effects), with the demo paused: dancer Off 1.0, Calm 2.0, Up to 3 per second 1.0 (clip 1.0); plaza 1.0, 1.0, 1.0 (clip 0.5); Every 1 on the dancer 1.0, 1.5, 1.5. Each within its setting's rule (the plaza's Off at the half-flash margin). Relight inside flares through its own gate, which the showreel's Flashing sets (its review A2). 0.1.28 (twenty-seven effects), with the demo paused: dancer Off 1.0, Calm 1.5, Up to 3 per second 1.0 (clip 1.0), red 0, within the bars. 0.1.29: not re-measured. Particles shows Emit here, whose output is unchanged (its review, A2). + dev/flash.py + dev/flashcount.py on the 0.1.0 build with the six effects inside (particles and point_cloud getting the dancer's depth, Depth Anything 3 Base, on input 3, showing the clip's frame), one run at a time, stress beat of 8 beats, 2 bars and 1 drop a second (Cycle then moves every 2 s, the fastest it allows). Worst second, area rule: Cycle on the dancer Off 1.0, Calm 2.0, Up to 3 per second 3.0 (clip alone 1.0); on the plaza 0.5, 1.5, 3.0 (clip 0.5); at Every 1 on the dancer 1.0, 2.0, 2.5. Cycle's cut, before the fade, reached 3.5 at Up to 3 per second with six effects (FAIL); the 1 s fade fixed it. No red flashes. All within flashcount's bars. 0.1.1 (Datamosh's bloom and Particles' sparks gated, both new): Cycle on the dancer Off 1.0, Calm 2.0, Up to 3 per second 2.5 (clip 1.0); plaza 0.5, 1.5, 3.0 (clip 0.5); Every 1 on the dancer 1.0, 2.0, 2.5. No red flashes; each within its setting's rule. 0.1.2 (the new dancer depth, Particles 0.3.1, Point Cloud 0.1.2): Cycle on the dancer Off 1.0, Calm 2.0, Up to 3 per second 3.0 (clip 1.0), within WCAG's 3 as 0.1.0 was; the plaza carries no depth, so its numbers stand. 0.1.3 (Fluid added): Cycle on the dancer Off 1.0, Calm 2.0, Up to 3 per second 3.0 (clip 1.0), red 0.5 at Up to 3 per second (Glitch's color split, which measures 1.0 red on its own, happened to show in the 10 s; under WCAG's 3); plaza 0.5, 1.5, 2.5 (clip 0.5); Every 1 on the dancer 1.0, 2.0, 2.5. Each within its setting's rule. 0.1.4 (VHS + CRT added): Cycle on the dancer Off 1.0, Calm 2.0, Up to 3 per second 2.0 (clip 1.0); plaza 0.5, 1.5, 1.5 (clip 0.5); Every 1 on the dancer 1.0, 2.0, 2.5; red 0.5 at Up to 3 per second on the dancer and the plaza (Glitch's split), none otherwise. Each within its setting's rule. 0.1.5 (VHS + CRT 0.2.0, every effect rebuilt for the lib change): Cycle on the dancer Off 1.5, Calm 2.0, Up to 3 per second 2.5 (clip 1.0; Off within flashcount's 0.5 of measuring noise, which effects showed in the 10 s differs from 0.1.4's 1.0); plaza 0.5, 1.5, 1.0 (clip 0.5); Every 1 on the dancer 1.0, 2.0, 2.5; red 0.5 at Up to 3 per second on all three (Glitch's split), none otherwise. Each within its setting's rule. 0.1.6: not re-measured: only the shader clock changed, wrapped every 16384 s (board #4681), which changes its value, not what the effect does with it, apart from one jump each 4.6 h. 0.1.7 (Bent Cam added, every effect rebuilt for the lib change): Cycle on the dancer Off 1.0, Calm 1.5, Up to 3 per second 2.5 (clip 1.0); plaza 1.0, 1.5, 2.0 (clip 0.5; Off within flashcount's 0.5 of measuring noise); Every 1 on the dancer 1.0, 1.5, 2.5; red 1.0 at Up to 3 per second on the dancer and 0.5 on the plaza (Glitch's split, the only effect with red flashes on the dancer: Glitch alone measures 1.0 at Calm and Up to 3), none otherwise. Each within its setting's rule. 0.1.8 (Bent Cam 0.1.1 inside): Cycle on the dancer Off 1.0, Calm 1.5, Up to 3 per second 2.5 (clip 1.0); red 1.0 at Up to 3 per second (Glitch's split), none otherwise. Each within its setting's rule. The plaza and Every 1 not re-measured: only Bent Cam's beats changed, measured on its own page. 0.1.9 (Slit-Scan added): before the fix, Cycle on the dancer measured Up to 3 per second 3.5 (FAIL): a fade from Bent Cam into Blob Track with Blob Track flashing as fast as Up to 3 allows in the same second, a pairing the 10 s window hadn't caught before. Now Up to 3 per second holds the effects at Calm while Cycle fades and for a second after. After: dancer Off 1.0, Calm 1.5, Up to 3 per second 2.0 (clip 1.0); plaza 1.0, 1.0, 2.0 (clip 0.5; Off within flashcount's 0.5 of measuring noise); Every 1 on the dancer 1.0, 1.5, 2.0. No red flashes. Each within its setting's rule. 0.1.10 (eleven effects), on the final build: dancer Off 1.0, Calm 1.5, Up to 3 per second 2.5 (clip 1.0); plaza 1.0, 1.0, 1.5 (clip 0.5; Off within flashcount's 0.5 of measuring noise); Every 1 on the dancer 1.0, 1.5, 2.0. No red flashes. Each within its setting's rule. 0.1.11 (twelve effects), on the final build with the demo paused: dancer Off 1.0, Calm 1.5, Up to 3 per second 2.0 (clip 1.0); plaza 0.5, 1.0, 1.5 (clip 0.5); Every 1 on the dancer 1.0, 1.0, 2.0. No red flashes. Each within its setting's rule. 0.1.12 embeds every effect as rebuilt with its network laid out (Network Boxes, positions), the unused compute shaders, and lays out its own network the same way; the README holds one line per paragraph. dev/probes/structure_diff.py against 0.1.11 finds every other operator, parameter, expression and wire unchanged, so these numbers stand. 0.1.13 (thirteen effects), on the final build with the demo paused: dancer Off 1.0, Calm 1.5, Up to 3 per second 2.0 (clip 1.0); plaza 1.0, 1.0, 2.0 (clip 0.5; Off within flashcount's 0.5 of measuring noise); Every 1 on the dancer 1.0, 1.5, 2.0. No red flashes. Each within its setting's rule. 0.1.14 (fourteen effects), on the final build with the demo paused: dancer Off 1.0, Calm 1.5, Up to 3 per second 1.5 (clip 1.0); plaza 0.5, 1.0, 1.5 (clip 0.5); Every 1 on the dancer 1.0, 1.0, 1.5. No red flashes. Each within its setting's rule. 0.1.15 (fifteen effects), on the final build with the demo paused: dancer Off 1.0, Calm 1.5, Up to 3 per second 1.5 (clip 1.0); plaza 1.0, 1.0, 1.5 (clip 0.5; Off within flashcount's 0.5 of measuring noise); Every 1 on the dancer 1.0, 1.5, 1.5. No red flashes. Each within its setting's rule. 0.1.16 (sixteen effects), on the final build with the demo paused: dancer Off 1.0, Calm 1.5, Up to 3 per second 1.5 (clip 1.0); plaza 1.0, 0.5, 1.5 (clip 0.5; Off within flashcount's 0.5 of measuring noise); Every 1 on the dancer 1.0, 1.5, 1.5. No red flashes. Each within its setting's rule. Video Vocoder inside keeps its own Flashing (its review A2). 0.1.17 (seventeen effects), with the demo paused: dancer Off 1.0, Calm 1.0, Up to 3 per second 1.5 (clip 1.0); plaza 1.0, 0.5, 1.0 (clip 0.5; Off within flashcount's 0.5 of measuring noise); Every 1 on the dancer 1.0, 1.5, 1.0. No red flashes. Each within its setting's rule. Detail inside keeps its own Flashing (its review A2). 0.1.18 (eighteen effects), with the demo paused: dancer Off 1.0, Calm 2.0, Up to 3 per second 2.0 (clip 1.0; Calm at its full allowance); plaza 0.5, 1.0, 1.0 (clip 0.5); Every 1 on the dancer 1.0, 1.5, 1.5. No red flashes. Each within its setting's rule. Chladni Sand inside keeps its own Flashing (its review A2). 0.1.19 (nineteen effects), with the demo paused: dancer Off 1.0, Calm 1.5, Up to 3 per second 1.5 (clip 1.0); plaza 0.5, 1.0, 1.0 (clip 0.5); Every 1 on the dancer 1.0, 1.0, 1.5. No red flashes. Each within its setting's rule. Pixel Art inside keeps its own Flashing (its review A2). 0.1.20 (twenty effects), with the demo paused: dancer Off 1.0, Calm 2.0, Up to 3 per second 1.5 (clip 1.0); plaza 1.0, 1.0, 1.0 (clip 0.5); Every 1 on the dancer 1.0, 1.5, 1.5. No red flashes. Each within its setting's rule. Lens Flare inside keeps its own Flashing (its review A2). 0.1.21 (twenty-one effects), with the demo paused: dancer Off 1.0, Calm 2.0, Up to 3 per second 1.5 (clip 1.0); plaza 1.0, 1.0, 1.0 (clip 0.5); Every 1 on the dancer 1.0, 1.5, 1.5. No red flashes. Each within its setting's rule. Words inside keeps its own Flashing (its review A2). 0.1.22 (twenty-two effects), with the demo paused: dancer Off 1.0, Calm 2.0, Up to 3 per second 1.5 (clip 1.0); plaza 1.0, 1.0, 1.0 (clip 0.5); Every 1 on the dancer 1.0, 1.5 (red 0.5), 1.0. Each within its setting's rule (the plaza's Off at the half-flash margin). Scan Processor inside keeps its own Flashing (its review A2). 0.1.23 (twenty-three effects), with the demo paused: dancer Off 1.0, Calm 1.5, Up to 3 per second 1.5 (clip 1.0); plaza 1.0, 1.0, 1.0 (clip 0.5); Every 1 on the dancer 1.0, 1.0, 1.5. Each within its setting's rule (the plaza's Off at the half-flash margin). Voxel Space inside keeps its own Flashing (its review A2). 0.1.24 (twenty-four effects), with the demo paused: dancer Off 1.0, Calm 1.5, Up to 3 per second 1.5 (clip 1.0); plaza 1.0, 1.0, 1.0 (clip 0.5); Every 1 on the dancer 1.0, 1.5, 1.5. Each within its setting's rule (the plaza's Off at the half-flash margin). Temporal Resonator inside has no Flashing setting and adds nothing to its clips' own (its review A2). 0.1.25 (twenty-five effects), with the demo paused: dancer Off 1.0, Calm 2.0, Up to 3 per second 1.5 (clip 1.0); plaza 1.0, 1.0, 1.0 (clip 0.5); Every 1 on the dancer 1.0, 1.5, 1.0. Each within its setting's rule (the plaza's Off at the half-flash margin). Caustics inside has no Flashing setting and adds nothing to its clips' own beyond the margin (its review A2). 0.1.26: not re-measured. The passes Caustics and Temporal Resonator left out never cooked (their reviews, C5), so the output is the same. 0.1.27 (twenty-six effects), with the demo paused: dancer Off 1.0, Calm 2.0, Up to 3 per second 1.0 (clip 1.0); plaza 1.0, 1.0, 1.0 (clip 0.5); Every 1 on the dancer 1.0, 1.5, 1.5. Each within its setting's rule (the plaza's Off at the half-flash margin). Relight inside flares through its own gate, which the showreel's Flashing sets (its review A2). 0.1.28 (twenty-seven effects), with the demo paused: dancer Off 1.0, Calm 1.5, Up to 3 per second 1.0 (clip 1.0), red 0, within the bars. 0.1.29: not re-measured. Particles shows Emit here, whose output is unchanged (its review, A2). 0.1.30: not re-measured. The output is unchanged: only shader comments changed inside. - [x] A3 No meaning rests on hue alone (brightness, size or motion carries it too); the palette is colorblind-safe (orange and blue); text is legible at the size it is shown. - The names are white (0.9) on a 60% dark plate, so they read by brightness on light and dark pictures, in the 5 x 7 font at 3 px per font pixel (at 720 high); checked on saved frames across a fade (dev/probes/showreel_fade.py). No meaning rests on color; the effects inside keep their reviewed palettes. 0.1.3: Fluid's label is FLUID in the same font. 0.1.4: VHS + CRT's label is VHS + CRT in the same font (the font has '+'). 0.1.5: unchanged. 0.1.7: Bent Cam's label is BENT CAM in the same font. 0.1.8: unchanged. 0.1.9: Slit-Scan's label is SLIT-SCAN in the same font (the font has '-'). 0.1.10: Pixel Sort's label is PIXEL SORT in the same font. 0.1.11: Paint's label is PAINT in the same font. 0.1.13: Reaction-Diffusion's label is REACTION-DIFFUSION in the same font. 0.1.14: Pencil's label is PENCIL in the same font. 0.1.15: Motion Magnify's label is MOTION MAGNIFY in the same font. 0.1.16: Video Vocoder's label is VIDEO VOCODER in the same font. 0.1.17: Detail's label is DETAIL in the same font. 0.1.18: Chladni Sand's label is CHLADNI SAND in the same font. 0.1.19: Pixel Art's label is PIXEL ART in the same font. 0.1.20: Lens Flare's label is LENS FLARE in the same font. 0.1.21: Words' label is WORDS in the same font. 0.1.22: Scan Processor's label is SCAN PROCESSOR in the same font. 0.1.23: Voxel Space's label is VOXEL SPACE in the same font. 0.1.24: Temporal Resonator's label is TEMPORAL RESONATOR in the same font. 0.1.25: Caustics' label is CAUSTICS in the same font. 0.1.27: Relight's label is RELIGHT in the same font. + The names are white (0.9) on a 60% dark plate, so they read by brightness on light and dark pictures, in the 5 x 7 font at 3 px per font pixel (at 720 high); checked on saved frames across a fade (dev/probes/showreel_fade.py). No meaning rests on color; the effects inside keep their reviewed palettes. 0.1.3: Fluid's label is FLUID in the same font. 0.1.4: VHS + CRT's label is VHS + CRT in the same font (the font has '+'). 0.1.5: unchanged. 0.1.7: Bent Cam's label is BENT CAM in the same font. 0.1.8: unchanged. 0.1.9: Slit-Scan's label is SLIT-SCAN in the same font (the font has '-'). 0.1.10: Pixel Sort's label is PIXEL SORT in the same font. 0.1.11: Paint's label is PAINT in the same font. 0.1.13: Reaction-Diffusion's label is REACTION-DIFFUSION in the same font. 0.1.14: Pencil's label is PENCIL in the same font. 0.1.15: Motion Magnify's label is MOTION MAGNIFY in the same font. 0.1.16: Video Vocoder's label is VIDEO VOCODER in the same font. 0.1.17: Detail's label is DETAIL in the same font. 0.1.18: Chladni Sand's label is CHLADNI SAND in the same font. 0.1.19: Pixel Art's label is PIXEL ART in the same font. 0.1.20: Lens Flare's label is LENS FLARE in the same font. 0.1.21: Words' label is WORDS in the same font. 0.1.22: Scan Processor's label is SCAN PROCESSOR in the same font. 0.1.23: Voxel Space's label is VOXEL SPACE in the same font. 0.1.24: Temporal Resonator's label is TEMPORAL RESONATOR in the same font. 0.1.25: Caustics' label is CAUSTICS in the same font. 0.1.27: Relight's label is RELIGHT in the same font. 0.1.30: unchanged. Every effect inside is rebuilt for lib's new parts (warmth(), incompressible(), spectrum.glsl); dev/probes/structure_diff.py finds only shader text changed (the headers of Fluid's divergence, pressure and velocity, and Lens Flare's burst with the CIE fit now from lib). - [x] A4 Anything unreliable is labeled experimental in its help and docs. - Nothing of the showreel's own is unreliable; the experimental parts of the effects inside (Fosfora's drops, the raw-audio kick detector) are labeled in their own help and pages. 0.1.1: the effects now describe raw-audio beats as a kick detector and a beat clock, experimental, in their help and pages. 0.1.3: Fluid's experimental parts (Fosfora's drops, raw-audio beats, the cut heuristic) are labeled in its own help and pages. 0.1.4: VHS + CRT's experimental parts (Fosfora's drops, raw-audio beats) are labeled in its own help and pages. 0.1.5: unchanged; VHS + CRT 0.2.0's experimental parts are labeled on its own page. 0.1.7: Bent Cam's experimental parts (Fosfora's drops, raw-audio beats) are labeled on its own page and in its help. 0.1.8: unchanged. 0.1.9: Slit-Scan's experimental parts (Fosfora's drops, raw-audio beats, Flow at occlusions) are labeled in its own help and page. 0.1.10: Pixel Sort's experimental parts (Fosfora's drops, raw-audio beats) are labeled in its own help and page. 0.1.11: Paint's experimental parts (Fosfora's drops, raw-audio beats) are labeled in its own help and page. 0.1.13: Reaction-Diffusion's experimental parts (Fosfora's drops, raw-audio beats) are labeled in its own help and page. 0.1.14: Pencil's experimental parts (Fosfora's drops, raw-audio beats) are labeled in its own help and page. 0.1.15: Motion Magnify's experimental parts (Fosfora's drops, raw-audio beats, where the optical flow goes wrong) are labeled in its own help and page. 0.1.16: Video Vocoder's experimental part (the bands Fosfora's bass, melody and drums stand in for without its feature bus) is labeled in its own README and page. 0.1.17: Detail has nothing experimental of its own (its review, A4). 0.1.18: Chladni Sand's experimental part (the bands Fosfora's bass, melody and drums stand in for without its feature bus) is labeled in its own README and page. 0.1.19: Pixel Art has nothing experimental of its own (its review, A4). 0.1.20: Lens Flare has nothing experimental of its own (its review, A4). 0.1.21: Words has nothing experimental of its own (its review, A4). 0.1.22: Scan Processor has nothing experimental of its own beyond audio_in()'s stand-ins for Fosfora's bands (its review, A4). 0.1.23: Voxel Space has nothing experimental of its own (its review, A4). 0.1.24: Temporal Resonator labels the cut that restarts its strings experimental (its review, A4). 0.1.25: Caustics labels the cut that keeps a new shot from slamming its water experimental (its review, A4). 0.1.27: Relight has nothing experimental of its own (its review, A4). + Nothing of the showreel's own is unreliable; the experimental parts of the effects inside (Fosfora's drops, the raw-audio kick detector) are labeled in their own help and pages. 0.1.1: the effects now describe raw-audio beats as a kick detector and a beat clock, experimental, in their help and pages. 0.1.3: Fluid's experimental parts (Fosfora's drops, raw-audio beats, the cut heuristic) are labeled in its own help and pages. 0.1.4: VHS + CRT's experimental parts (Fosfora's drops, raw-audio beats) are labeled in its own help and pages. 0.1.5: unchanged; VHS + CRT 0.2.0's experimental parts are labeled on its own page. 0.1.7: Bent Cam's experimental parts (Fosfora's drops, raw-audio beats) are labeled on its own page and in its help. 0.1.8: unchanged. 0.1.9: Slit-Scan's experimental parts (Fosfora's drops, raw-audio beats, Flow at occlusions) are labeled in its own help and page. 0.1.10: Pixel Sort's experimental parts (Fosfora's drops, raw-audio beats) are labeled in its own help and page. 0.1.11: Paint's experimental parts (Fosfora's drops, raw-audio beats) are labeled in its own help and page. 0.1.13: Reaction-Diffusion's experimental parts (Fosfora's drops, raw-audio beats) are labeled in its own help and page. 0.1.14: Pencil's experimental parts (Fosfora's drops, raw-audio beats) are labeled in its own help and page. 0.1.15: Motion Magnify's experimental parts (Fosfora's drops, raw-audio beats, where the optical flow goes wrong) are labeled in its own help and page. 0.1.16: Video Vocoder's experimental part (the bands Fosfora's bass, melody and drums stand in for without its feature bus) is labeled in its own README and page. 0.1.17: Detail has nothing experimental of its own (its review, A4). 0.1.18: Chladni Sand's experimental part (the bands Fosfora's bass, melody and drums stand in for without its feature bus) is labeled in its own README and page. 0.1.19: Pixel Art has nothing experimental of its own (its review, A4). 0.1.20: Lens Flare has nothing experimental of its own (its review, A4). 0.1.21: Words has nothing experimental of its own (its review, A4). 0.1.22: Scan Processor has nothing experimental of its own beyond audio_in()'s stand-ins for Fosfora's bands (its review, A4). 0.1.23: Voxel Space has nothing experimental of its own (its review, A4). 0.1.24: Temporal Resonator labels the cut that restarts its strings experimental (its review, A4). 0.1.25: Caustics labels the cut that keeps a new shot from slamming its water experimental (its review, A4). 0.1.27: Relight has nothing experimental of its own (its review, A4). 0.1.30: unchanged. Every effect inside is rebuilt for lib's new parts (warmth(), incompressible(), spectrum.glsl); dev/probes/structure_diff.py finds only shader text changed (the headers of Fluid's divergence, pressure and velocity, and Lens Flare's burst with the CIE fit now from lib). diff --git a/bundles/showreel/showreel.json b/bundles/showreel/showreel.json index 061f40c..b697b31 100644 --- a/bundles/showreel/showreel.json +++ b/bundles/showreel/showreel.json @@ -1,37 +1,37 @@ { - "version": "0.1.29", - "sources_sha256": "6d4a5aef5f82052b4a385340974589d26f4321980fd3ddf5574cbcd354298b3f", + "version": "0.1.30", + "sources_sha256": "3b3ab98d2536f47a09cf615735c0da70e2453c9ac59bfe94d71a0e4988a2579d", "embeds": { - "ascii": "301ff3ad49c284f4dfc5fd9cce9312ae6d864860f1a0b580aa4858f51177d117", - "bent_cam": "26f433b2d444873d7c428b4a6899a4ee01bb204b73032761058fa82b23c27998", - "blob_track": "656532265a7abfa48cdbdfa5e42d622f7b8ec87787c2a060559a089a02a73081", - "caustics": "6dcadb22c9eed699c4a8512bc41197dd9a1b81ce32c73ef3f1df6a4aacba3a93", - "chladni_sand": "b45d5ee105959897b754533ab3404b6242232665335b3fa48f33ef29488b99bf", - "datamosh": "c6acd82a9ebb9389c3c72da22dc5338c3e8fddadebf4ca97532f6250673fbc75", - "detail": "04720c8d523e2440d939f0f5fa2c35eea3fd31fe05c4b456d70c8961055285bf", - "film_stock": "519de6e07a3c6dadf3e4abb14e3b54dda97288a6bc785d989317078b22fb9674", - "fluid": "ff31a838695606d16f7e879ddab684c0e79bee9286b748acea38d2b3c98b8a92", - "glitch": "2fbc8aae2f4cd41408459655de7f8c6922a276b542476b898adeddac10c9e714", - "lens_flare": "4aa730780afc53ed4b736a16ff1be92e5793cb9f522aaf1e4346d59942958996", - "motion_magnify": "c7f8b00d98fee30b0f9de1de7b3bc7f4779ce341d6072fb892ef293dde9df24f", - "paint": "8cf47f0725e5b54130f3070f216b65b0a71ab5ef3335196f575058fe45f9844e", - "particles": "c2abacfb3a52b0498d73d78572e0faf1a4bc1089d97f4e9c5cc1f0b10d3ccc68", - "pencil": "bb3bc3b0c37074510ec1a0db739b538528c63c3860538c6353a3e4afe481281e", - "pixel_art": "33168e9ae605f14e2eb4c3b6fe8b936fd8a61511c9a01f4772f41a3eb173f0b8", - "pixel_sort": "50d3845fb9d7b8c9f72470af1bdcafede16c1c8a6ef5f0db7f08f52b1f4ca4de", - "point_cloud": "f152ec7d0670ffb22690aaa40b7f81b27aa2b5138de49aa7cf20b66a49edca5a", - "reaction_diffusion": "f33f37525dd1a5189879563b1afb213bd07bb74794e23d5b4a5f3886c56abd5d", - "relight": "499c4c373315d3a6224dd30247ae2751b4fac5158c9aa3b0986087844de34039", - "scan_processor": "2c82d8bba962051f12fd31a369bf0f77e439fc4e1b4b7d5dacba12d31eced5d1", - "slit_scan": "e933057bdde9dc560d8e6038266e89a89f2856e5e73ecb760f7501ede87f0300", - "temporal_resonator": "67ef0634916acdb9f29d76717ae15f698ccc09579ca2da116ba56dffe840e98f", - "vhs_crt": "5e49ec255ed696274f0673077196c12e4772eda18589a82c0885c6e2d27a70f7", - "video_vocoder": "6ed9293a0fb8a7ab0aae4b4cdc96a92a1c1f0b81e072a91835c276346ac70316", - "voxel_space": "09249a79a4f0b821df9be81cae6625b68493b605ef5331fd6d1d5d3fc6defdce", - "words": "e54254f20615392d4155cab16417cb7c754f400c4e5f9396123ad18196d30b5c" + "ascii": "c9eb845e4072e4d17eeab10b94baa25a3bd54bb3938fb0f629a23b1474788753", + "bent_cam": "289d490a09d173b3e2bd3dc2fac0f1c025edbb335dae5d55deb388c875e20e45", + "blob_track": "9104d978eac9190f8c77d171f1535b3a643509b357da49eab5b4ae513195a3b8", + "caustics": "3aafa144a12e205991caee72661b0c3a76b3b45d7e2f88c2c8f8d34c8d4abde8", + "chladni_sand": "4d620bc18d3aafe1f7f208ddb952d038b7cd1f0fbca0e8e33644da809a06e29d", + "datamosh": "2c165c28717cf35b70982d0462fa0b9935fa9ee7834b8b33f38c5ee6514163b4", + "detail": "10127cf631276692d9bc8941e4361d77791f54fc31277a718b2ef6d2404f077a", + "film_stock": "bc1766177da183442e68a41d0e43871d4f99e15a7b7d4845fd9a2d386ed55f90", + "fluid": "968bdd0c0744bbdcf77e8f65cf636aad96fccf10411d36e7e738efe0b0645b64", + "glitch": "3d62d2af636bd90284cf307074ad32453a40919ebc431da3baa73b975ce86771", + "lens_flare": "fe986bfa3a3103345a237d2c74da5570c0ebd6f1d6500ef3547a3091570512eb", + "motion_magnify": "3ffd44c6c05021e84943f424d437da96b5040715be903608b126188d9bc31e31", + "paint": "c5fa558601d592cc465e2cf6d1ca6563000049f4936774908ac50edf594c17fe", + "particles": "b802ba91fa0b33e26f6daa514531734eba8d2d346d2491c634cdd2af5ac4a77b", + "pencil": "8a56bd9261704153d85edf1cb30fec108a1b2496330e47b4febc8cb83eae753b", + "pixel_art": "6b88e6150105dead868e9dfb8f39d4aa5c28801085a85e6159483a764cecb7b0", + "pixel_sort": "0dea72623967614b64d064100b900ed9d5c1f2b84000f17dde8a28e40195d5ff", + "point_cloud": "55873bf27c8e61f66823bd3dfb5e2a1fa34a8a4022739bb25613628fc74ac0d9", + "reaction_diffusion": "2eb1edebc60193dfcc1df3fd9538506b4462039fc7c454e4a989ae39dd711b50", + "relight": "d701cf632971d931646854a4b2c403e72daca453248f66d8674b18b2da1d306f", + "scan_processor": "62f0c70ba957b1ff2a4a199559389dabc4499a99f09ba302c1a28a887f103712", + "slit_scan": "0c98addf816a35bbc8ff6d5764962c140e71af764ec84a8ee72a59d33b9b9864", + "temporal_resonator": "f1789c87acaecb26f24a51dda157ddc2829e700541e70ca1cd29b51aabff75f8", + "vhs_crt": "bc99c4240e3fb32d180abe424431f0d964a90ba2524fe5afef82b74e13763849", + "video_vocoder": "7a4620724537d176d9dbeb13da4babd8ac356719bfa23f82cce972e923293ab7", + "voxel_space": "1c1ff829b08406067efa172bae39b4e2ea8eb9d55919fd2942c67dfcc571e31e", + "words": "21c70c56305c43bf28a3e1db396326d38dd05ebe3111199ade3a0da0847ada04" }, "embeds_all": true, "touchdesigner_build": "2025.33230", - "tox_sha256": "0262b151e1843acf0e95d89c2c5fc6fdf5dc03617dd85b47129033c346329589", - "tox_bytes": 3147264 + "tox_sha256": "df2c82e10c4f6aa7d0c9f4128f69ca04826326df231dc7a73aff23470be90428", + "tox_bytes": 3147304 } diff --git a/bundles/showreel/showreel.tox b/bundles/showreel/showreel.tox index 698168e..5412125 100644 Binary files a/bundles/showreel/showreel.tox and b/bundles/showreel/showreel.tox differ diff --git a/docs/BUILDING.md b/docs/BUILDING.md index 9837724..6f637cc 100644 --- a/docs/BUILDING.md +++ b/docs/BUILDING.md @@ -17,7 +17,10 @@ lib/glsl/ shaders several effects share: depth_near.glsl (in red_guard.glsl (keeps reds short of WCAG's red flash line); tensor, tensor_blur, tensor_ease and tensor_flow (the direction of the picture's edges: structure_tensor()); paper.glsl (gradient noise and a paper's grain, drawn rather than loaded: Paint, Pencil); - pyramid_down.glsl and pyramid.glsl (the picture's octaves of detail: pyramid()) + pyramid_down.glsl and pyramid.glsl (the picture's octaves of detail: pyramid()); + divergence, pressure and project (a fluid's velocity made incompressible: + incompressible()); warmth.glsl (what the picture warms: what stands out and what + moves, warmth()); spectrum.glsl (the CIE color matching functions: Lens Flare) effects// build_.py the builder (has a VERSION = 'x.y.z' line) src/ what it is built from: *.glsl, *.py @@ -104,6 +107,11 @@ the size of the one before (Video Vocoder, Detail): a shader that pastes `lib_gl Laplacian band as a level minus `expand()` of the next, and summing the bands back up through the same `expand()` gives the picture back exactly (within a fraction of an 8-bit step at 32-bit float). An effect with several pyramids names each one's levels with `prefix=` (Detail's `l1`..., `a1`..., `b1`...) and says what it halves with `what=`; +`incompressible(comp, _ROOT, advected, passes)` makes a fluid's velocity incompressible, the pressure solve of stable +fluids (divergence, then `passes` Jacobi iterations of the pressure in one TOP, then the projection), as Fluid runs it; +`warmth(comp, _ROOT, subject, mo, start)` blends what stands out or is inside the mask (subject) and what moves +(`optical_flow()`'s change and clock, held between the input's frames, past a dead zone) into one field that rises +quickly and falls slowly, which Frost thaws by; `lib_module(_ROOT, 'font5x7')` loads another `lib/` file, such as the bitmap font (`glsl_font()` gives the shader `glyph(g, p)`; `ramp()` orders glyphs by ink), and `lib_glsl(_ROOT, 'depth_near.glsl')` a shared shader. diff --git a/effects/ascii/ascii.json b/effects/ascii/ascii.json index 1c1c78e..43c4acc 100644 --- a/effects/ascii/ascii.json +++ b/effects/ascii/ascii.json @@ -1,11 +1,11 @@ { "version": "0.2.1", - "sources_sha256": "3c024e5d2d13f5d7c173eaccdf984d02971ca405960e0d9855efb31aed3808aa", + "sources_sha256": "d911daa50e984e0a1d70bddc748279fe786d3852e9902c4118014abcf11664b2", "inputs": [ "in1_top", "in2_audio" ], "touchdesigner_build": "2025.33230", - "tox_sha256": "301ff3ad49c284f4dfc5fd9cce9312ae6d864860f1a0b580aa4858f51177d117", + "tox_sha256": "c9eb845e4072e4d17eeab10b94baa25a3bd54bb3938fb0f629a23b1474788753", "tox_bytes": 32534 } diff --git a/effects/ascii/ascii.tox b/effects/ascii/ascii.tox index 02d009c..6f6ea51 100644 Binary files a/effects/ascii/ascii.tox and b/effects/ascii/ascii.tox differ diff --git a/effects/bent_cam/bent_cam.json b/effects/bent_cam/bent_cam.json index 53c346c..6b5b57c 100644 --- a/effects/bent_cam/bent_cam.json +++ b/effects/bent_cam/bent_cam.json @@ -1,11 +1,11 @@ { "version": "0.1.2", - "sources_sha256": "deb79a7c29f1c8222237d25d6f1bbfda2ddc837a6fb51959597d4d35e3232aa4", + "sources_sha256": "d87404d0ec2cc7a34573721307d35097e43ac69aa388c36a20c9eadd8741605f", "inputs": [ "in1_top", "in2_audio" ], "touchdesigner_build": "2025.33230", - "tox_sha256": "26f433b2d444873d7c428b4a6899a4ee01bb204b73032761058fa82b23c27998", + "tox_sha256": "289d490a09d173b3e2bd3dc2fac0f1c025edbb335dae5d55deb388c875e20e45", "tox_bytes": 51950 } diff --git a/effects/bent_cam/bent_cam.tox b/effects/bent_cam/bent_cam.tox index 07ed70a..1adf330 100644 Binary files a/effects/bent_cam/bent_cam.tox and b/effects/bent_cam/bent_cam.tox differ diff --git a/effects/blob_track/blob_track.json b/effects/blob_track/blob_track.json index e688099..af8bc84 100644 --- a/effects/blob_track/blob_track.json +++ b/effects/blob_track/blob_track.json @@ -1,11 +1,11 @@ { "version": "0.1.4", - "sources_sha256": "bf3086068433f139ac3d2d6564021a252a8454bbd7b1522772e08e4d2eb6471d", + "sources_sha256": "ec4d9406d407959bc66c99c6bac15571e5c3cc41bff9a55e1a42a82bfc6dc6db", "inputs": [ "in1_top", "in2_audio" ], "touchdesigner_build": "2025.33230", - "tox_sha256": "656532265a7abfa48cdbdfa5e42d622f7b8ec87787c2a060559a089a02a73081", + "tox_sha256": "9104d978eac9190f8c77d171f1535b3a643509b357da49eab5b4ae513195a3b8", "tox_bytes": 52774 } diff --git a/effects/blob_track/blob_track.tox b/effects/blob_track/blob_track.tox index 7ab73b8..7c05f0e 100644 Binary files a/effects/blob_track/blob_track.tox and b/effects/blob_track/blob_track.tox differ diff --git a/effects/caustics/caustics.json b/effects/caustics/caustics.json index c118f42..4e94f89 100644 --- a/effects/caustics/caustics.json +++ b/effects/caustics/caustics.json @@ -1,12 +1,12 @@ { "version": "0.1.1", - "sources_sha256": "42f6cbf5adfa0f696aab4aadaba5c6f5498330c9f13844d3383b7a408aa4be30", + "sources_sha256": "99f50dc20161b13c1f4bb6dcb1240a3c820fdbb057ef55ad634b5891fe516abe", "inputs": [ "in1_top", "in2_audio", "in3_top" ], "touchdesigner_build": "2025.33230", - "tox_sha256": "6dcadb22c9eed699c4a8512bc41197dd9a1b81ce32c73ef3f1df6a4aacba3a93", + "tox_sha256": "3aafa144a12e205991caee72661b0c3a76b3b45d7e2f88c2c8f8d34c8d4abde8", "tox_bytes": 32878 } diff --git a/effects/caustics/caustics.tox b/effects/caustics/caustics.tox index f675896..d99770d 100644 Binary files a/effects/caustics/caustics.tox and b/effects/caustics/caustics.tox differ diff --git a/effects/chladni_sand/chladni_sand.json b/effects/chladni_sand/chladni_sand.json index 30bd1f1..6f2fde1 100644 --- a/effects/chladni_sand/chladni_sand.json +++ b/effects/chladni_sand/chladni_sand.json @@ -1,12 +1,12 @@ { "version": "0.1.0", - "sources_sha256": "70b0cf1be2db0dd00c425738935de4a1d010572e43b18febdb5a689f7bea6f9c", + "sources_sha256": "f1552f5cb87dd71f78cd150d75ab639e2271e2702b1ca4d647400074593d4625", "inputs": [ "in1_top", "in2_audio", "in3_top" ], "touchdesigner_build": "2025.33230", - "tox_sha256": "b45d5ee105959897b754533ab3404b6242232665335b3fa48f33ef29488b99bf", + "tox_sha256": "4d620bc18d3aafe1f7f208ddb952d038b7cd1f0fbca0e8e33644da809a06e29d", "tox_bytes": 27726 } diff --git a/effects/chladni_sand/chladni_sand.tox b/effects/chladni_sand/chladni_sand.tox index d35c107..0141223 100644 Binary files a/effects/chladni_sand/chladni_sand.tox and b/effects/chladni_sand/chladni_sand.tox differ diff --git a/effects/datamosh/datamosh.json b/effects/datamosh/datamosh.json index a3a1138..2d6eb5b 100644 --- a/effects/datamosh/datamosh.json +++ b/effects/datamosh/datamosh.json @@ -1,11 +1,11 @@ { "version": "0.2.2", - "sources_sha256": "4b09aaca84e2287e244ada7fdae9380abd5ba58ca47e5a466f5cc1e564b2d8ee", + "sources_sha256": "1a7170f931ab8f7229a4bb787232343ad98e82142d8341b8fe465c4eb5349054", "inputs": [ "in1_top", "in2_audio" ], "touchdesigner_build": "2025.33230", - "tox_sha256": "c6acd82a9ebb9389c3c72da22dc5338c3e8fddadebf4ca97532f6250673fbc75", + "tox_sha256": "2c165c28717cf35b70982d0462fa0b9935fa9ee7834b8b33f38c5ee6514163b4", "tox_bytes": 48366 } diff --git a/effects/datamosh/datamosh.tox b/effects/datamosh/datamosh.tox index ea6ae3e..c6e8fad 100644 Binary files a/effects/datamosh/datamosh.tox and b/effects/datamosh/datamosh.tox differ diff --git a/effects/detail/detail.json b/effects/detail/detail.json index 84eeb60..e14f1d5 100644 --- a/effects/detail/detail.json +++ b/effects/detail/detail.json @@ -1,12 +1,12 @@ { "version": "0.1.0", - "sources_sha256": "1a00f6f01414e3bd44a47cbc563b851270e2a638fdce336d9689dea0fd1c14f4", + "sources_sha256": "08303a07d712999966e60bdfc0dc3f9f67a934fc2cff8b9d3f8c4e28fb529fda", "inputs": [ "in1_top", "in2_audio", "in3_top" ], "touchdesigner_build": "2025.33230", - "tox_sha256": "04720c8d523e2440d939f0f5fa2c35eea3fd31fe05c4b456d70c8961055285bf", + "tox_sha256": "10127cf631276692d9bc8941e4361d77791f54fc31277a718b2ef6d2404f077a", "tox_bytes": 52358 } diff --git a/effects/detail/detail.tox b/effects/detail/detail.tox index f4c0d76..fa44eb2 100644 Binary files a/effects/detail/detail.tox and b/effects/detail/detail.tox differ diff --git a/effects/film_stock/film_stock.json b/effects/film_stock/film_stock.json index b9d995d..eff42ff 100644 --- a/effects/film_stock/film_stock.json +++ b/effects/film_stock/film_stock.json @@ -1,11 +1,11 @@ { "version": "0.1.0", - "sources_sha256": "cd6e00a558c8018ba33b0367f34b78a03ce7e656a833b44e637e4883e6128d59", + "sources_sha256": "3254c327301fc6be797fd99144f48dcff6808d284a6477bbb2ac99e9179ff3d2", "inputs": [ "in1_top", "in2_audio" ], "touchdesigner_build": "2025.33230", - "tox_sha256": "519de6e07a3c6dadf3e4abb14e3b54dda97288a6bc785d989317078b22fb9674", - "tox_bytes": 61422 + "tox_sha256": "bc1766177da183442e68a41d0e43871d4f99e15a7b7d4845fd9a2d386ed55f90", + "tox_bytes": 61414 } diff --git a/effects/film_stock/film_stock.tox b/effects/film_stock/film_stock.tox index 69d35bb..e5ea37d 100644 Binary files a/effects/film_stock/film_stock.tox and b/effects/film_stock/film_stock.tox differ diff --git a/effects/fluid/build_fluid.py b/effects/fluid/build_fluid.py index 3c4ab1c..6ce11a9 100644 --- a/effects/fluid/build_fluid.py +++ b/effects/fluid/build_fluid.py @@ -13,8 +13,8 @@ frame came, the gap between frames) -> motion (the flow as a speed, held between the input's frames) emit where the input pours into the fluid: motion, the picture, or input 3 (depth -> near -> find, or near alone for a mask; subject passes them only for Depth and Mask, so they cook only then) - fluid advect (the velocity carried by itself, swirled, risen and dragged by the motion) -> divergence - -> pressure (Jacobi, PASSES iterations in one TOP) -> velocity (made incompressible), read back + fluid advect (the velocity carried by itself, swirled, risen and dragged by the motion) -> lib incompressible(): + divergence -> pressure (Jacobi, PASSES iterations in one TOP) -> velocity (made incompressible), read back through velocity_fb; dye (what it carries, carried by velocity), read back through dye_fb; still is where every Feedback TOP starts, and Reset starts them again look the dye drawn over the picture as ink, smoke or flames @@ -240,18 +240,10 @@ def choices(p, items, default): "{})".format(SWIRL_K, per_look(SWIRL), CHURN, BEAT)), Expr('{} + parent().par.Rise + {} * {}'.format(per_look(RISE), LEAP, BEAT)), Expr(per_look(FLOW_HALF)))) -divergence = glsl(comp, 'divergence', [advect], src('divergence.glsl'), - 'How much fluid each cell gains or loses.', fmt='rgba32float') -pressure_fb = mk(comp, feedbackTOP, 'pressure_fb', [divergence], 'The pressure last frame.', top='pressure') -pressure = glsl(comp, 'pressure', [pressure_fb, divergence], src('pressure.glsl'), - 'The pressure that cancels the divergence: {} Jacobi iterations.'.format(PASSES), fmt='rgba32float') -setpars(pressure, npasses=PASSES) -velocity = glsl(comp, 'velocity', [advect, pressure], src('project.glsl'), - 'The velocity made incompressible: what the dye moves with.', fmt='rgba32float') +divergence, pressure, velocity = incompressible(comp, _ROOT, advect, PASSES) # Hold: a cell off the edge reads as the edge, so the velocity's differences don't jump there (the default, -# Zero, would). The pressure keeps Zero: the fluid flows out of the frame freely. -for g in (advect, divergence): - setpars(g, inputextenduv='hold') +# Zero, would; incompressible() holds the divergence's). The pressure keeps Zero: the fluid flows out freely. +setpars(advect, inputextenduv='hold') dye = glsl(comp, 'dye', [source, dye_fb, velocity, emit], lib_glsl(_ROOT, 'common.glsl') + src('dye.glsl'), 'What the fluid carries: ink, smoke or heat, carried by the velocity, poured in where the picture emits.', fmt='rgba16float', diff --git a/effects/fluid/fluid.json b/effects/fluid/fluid.json index ea94d2c..b221c2b 100644 --- a/effects/fluid/fluid.json +++ b/effects/fluid/fluid.json @@ -1,12 +1,12 @@ { "version": "0.1.2", - "sources_sha256": "787a9ece2d68684138c118d836e98ad5afe64c784e28a16ee5c2b3ae3225ce90", + "sources_sha256": "813a6ccc9b68c6954be1533986c1b74692c6c79dea6b7bf02732288ea3f46045", "inputs": [ "in1_top", "in2_audio", "in3_top" ], "touchdesigner_build": "2025.33230", - "tox_sha256": "ff31a838695606d16f7e879ddab684c0e79bee9286b748acea38d2b3c98b8a92", - "tox_bytes": 46126 + "tox_sha256": "968bdd0c0744bbdcf77e8f65cf636aad96fccf10411d36e7e738efe0b0645b64", + "tox_bytes": 46150 } diff --git a/effects/fluid/fluid.tox b/effects/fluid/fluid.tox index 5f82df0..ec8c439 100644 Binary files a/effects/fluid/fluid.tox and b/effects/fluid/fluid.tox differ diff --git a/effects/fluid/src/pressure.glsl b/effects/fluid/src/pressure.glsl deleted file mode 100644 index ea60eef..0000000 --- a/effects/fluid/src/pressure.glsl +++ /dev/null @@ -1,14 +0,0 @@ -// Fluid (touchdesigner-effects, MIT): the pressure that cancels the divergence, by Jacobi iteration -// of its Poisson equation (Harris 2004, GPU Gems ch. 38). The GLSL TOP runs this shader Passes times -// a frame, each pass reading the last one's result as input 0 (2025.33230, dev/probes/glsl_passes.py); -// the first pass starts from last frame's pressure, which is close, so a few dozen passes do. Off the -// frame's edges the pressure reads 0 (Input Extend Mode Zero): the fluid flows out freely. -// Inputs: 0 the pressure so far (last frame's on the first pass), 1 the divergence. Out: r the pressure. -out vec4 fragColor; -float p(vec2 q) { return texture(sTD2DInputs[0], q).r; } -void main() { - vec2 px = uTD2DInfos[0].res.xy; - vec2 uv = vUV.st; - float sum = p(uv + vec2(px.x, 0.0)) + p(uv - vec2(px.x, 0.0)) + p(uv + vec2(0.0, px.y)) + p(uv - vec2(0.0, px.y)); - fragColor = vec4(0.25 * (sum - texture(sTD2DInputs[1], uv).r), 0.0, 0.0, 1.0); -} diff --git a/effects/glitch/glitch.json b/effects/glitch/glitch.json index a0ff573..e0ae2c2 100644 --- a/effects/glitch/glitch.json +++ b/effects/glitch/glitch.json @@ -1,11 +1,11 @@ { "version": "0.1.2", - "sources_sha256": "f1bdeb3791f8767f154c0942ae137767a3040b58e022aab0b074cdfa9791e3a9", + "sources_sha256": "e0debcac3f250893a50276e430ddb7b80977d92cdaed56fff1934963456d37fe", "inputs": [ "in1_top", "in2_audio" ], "touchdesigner_build": "2025.33230", - "tox_sha256": "2fbc8aae2f4cd41408459655de7f8c6922a276b542476b898adeddac10c9e714", - "tox_bytes": 26158 + "tox_sha256": "3d62d2af636bd90284cf307074ad32453a40919ebc431da3baa73b975ce86771", + "tox_bytes": 26174 } diff --git a/effects/glitch/glitch.tox b/effects/glitch/glitch.tox index 92ce702..921ce76 100644 Binary files a/effects/glitch/glitch.tox and b/effects/glitch/glitch.tox differ diff --git a/effects/lens_flare/build_lens_flare.py b/effects/lens_flare/build_lens_flare.py index 9bea5d7..3655e28 100644 --- a/effects/lens_flare/build_lens_flare.py +++ b/effects/lens_flare/build_lens_flare.py @@ -227,7 +227,7 @@ def ghost_rows(spec): dft = glsl(comp, 'dft', [dft_rows], src('dft.glsl'), "The transform along the columns, squared: the iris's diffraction at one wavelength.", res=(N, N), fmt='r32float') -burst = glsl(comp, 'burst', [dft], shader('burst.glsl'), +burst = glsl(comp, 'burst', [dft], src('common.glsl') + lib_glsl(_ROOT, 'spectrum.glsl') + src('burst.glsl'), 'The starburst: the diffraction summed over 24 wavelengths, each scaled by its length.', res=(N, N), fmt='rgba16float', uBurst=(ZOOM, TONE)) setpars(burst, inputfiltertype='linear') diff --git a/effects/lens_flare/lens_flare.json b/effects/lens_flare/lens_flare.json index eaaba4b..93050df 100644 --- a/effects/lens_flare/lens_flare.json +++ b/effects/lens_flare/lens_flare.json @@ -1,11 +1,11 @@ { "version": "0.1.0", - "sources_sha256": "a85871675b8a9240dd062ec4c5f95db79705444467e86d43797f64ee8701347a", + "sources_sha256": "afc2f9a8d8deeb0acd3009468c217d4120ef1088a1cd6463db72cc00758feb66", "inputs": [ "in1_top", "in2_audio" ], "touchdesigner_build": "2025.33230", - "tox_sha256": "4aa730780afc53ed4b736a16ff1be92e5793cb9f522aaf1e4346d59942958996", - "tox_bytes": 61686 + "tox_sha256": "fe986bfa3a3103345a237d2c74da5570c0ebd6f1d6500ef3547a3091570512eb", + "tox_bytes": 61830 } diff --git a/effects/lens_flare/lens_flare.tox b/effects/lens_flare/lens_flare.tox index 52045b2..2d3ca96 100644 Binary files a/effects/lens_flare/lens_flare.tox and b/effects/lens_flare/lens_flare.tox differ diff --git a/effects/lens_flare/src/burst.glsl b/effects/lens_flare/src/burst.glsl index 56920fd..6806cc8 100644 --- a/effects/lens_flare/src/burst.glsl +++ b/effects/lens_flare/src/burst.glsl @@ -1,30 +1,13 @@ // touchdesigner-effects (MIT), Lens Flare: the starburst, a light's diffraction around the iris, after Hullin et al. // (2011). The iris's far-field pattern (dft.glsl) belongs to one wavelength; at another it is the same pattern scaled // by the wavelength, longer waves spreading wider. Summed over 24 wavelengths across the visible spectrum, each -// through the CIE 1931 color matching functions (C. Wyman, P.-P. Sloan and P. Shirley's analytic fit, "Simple -// Analytic Approximations to the CIE XYZ Color Matching Functions", JCGT 2(2), 2013) and into linear sRGB, the -// spikes come out white near the light and end in rainbows. A real light is far brighter than its clipped pixels -// show, so its faint spikes are seen: the brightness is raised to the power Tone (a quarter), and the pattern fades out -// toward the texture's edge. Cooks only when the iris changes. +// through the CIE 1931 color matching functions (lib spectrum.glsl: C. Wyman, P.-P. Sloan and P. Shirley's analytic +// fit, JCGT 2013) and into linear sRGB, the spikes come out white near the light and end in rainbows. A real light is +// far brighter than its clipped pixels show, so its faint spikes are seen: the brightness is raised to the power Tone +// (a quarter), and the pattern fades out toward the texture's edge. Cooks only when the iris changes. // Input: 0 the pattern at 550 nm. Out: the starburst, 1 at the middle. uniform vec4 uBurst; // x zoom: frequency steps per pixel at 550 nm; y Tone -float lobe(float lam, float mu, float s1, float s2) { - float t = (lam - mu) / (lam < mu ? s1 : s2); - return exp(-0.5 * t * t); -} - -vec3 cmf(float lam) { - return vec3(1.056 * lobe(lam, 599.8, 37.9, 31.0) + 0.362 * lobe(lam, 442.0, 16.0, 26.7) - - 0.065 * lobe(lam, 501.1, 20.4, 26.2), - 0.821 * lobe(lam, 568.8, 46.9, 40.5) + 0.286 * lobe(lam, 530.9, 16.3, 31.1), - 1.217 * lobe(lam, 437.0, 11.8, 36.0) + 0.681 * lobe(lam, 459.0, 26.0, 13.8)); -} - -const mat3 XYZ_TO_RGB = mat3(3.2404542, -0.9692660, 0.0556434, - -1.5371385, 1.8760108, -0.2040259, - -0.4985314, 0.0415560, 1.0572252); - out vec4 fragColor; void main() { vec2 n = vec2(textureSize(sTD2DInputs[0], 0)); diff --git a/effects/motion_magnify/motion_magnify.json b/effects/motion_magnify/motion_magnify.json index 3e12922..97effbb 100644 --- a/effects/motion_magnify/motion_magnify.json +++ b/effects/motion_magnify/motion_magnify.json @@ -1,12 +1,12 @@ { "version": "0.1.0", - "sources_sha256": "658815c36e0cc51b5e1065f1adcb25b7080c2e28713fa4d7858286b5e78fdd65", + "sources_sha256": "041e8e4ca16fdf731006088f2fab822b93cad3e13d4e2d496ec3f8a699563ab5", "inputs": [ "in1_top", "in2_audio", "in3_top" ], "touchdesigner_build": "2025.33230", - "tox_sha256": "c7f8b00d98fee30b0f9de1de7b3bc7f4779ce341d6072fb892ef293dde9df24f", + "tox_sha256": "3ffd44c6c05021e84943f424d437da96b5040715be903608b126188d9bc31e31", "tox_bytes": 47574 } diff --git a/effects/motion_magnify/motion_magnify.tox b/effects/motion_magnify/motion_magnify.tox index e349312..40ab3c2 100644 Binary files a/effects/motion_magnify/motion_magnify.tox and b/effects/motion_magnify/motion_magnify.tox differ diff --git a/effects/paint/paint.json b/effects/paint/paint.json index 904e97b..556b4c1 100644 --- a/effects/paint/paint.json +++ b/effects/paint/paint.json @@ -1,12 +1,12 @@ { "version": "0.1.1", - "sources_sha256": "26877775bf5c80141c1c8811336c71b71e03edf4ad7a9b00a1fab004cae5f124", + "sources_sha256": "10889234f2060429868b39d4b7527e7f866270eb950dbb7b4b2b87043bd13cae", "inputs": [ "in1_top", "in2_audio", "in3_top" ], "touchdesigner_build": "2025.33230", - "tox_sha256": "8cf47f0725e5b54130f3070f216b65b0a71ab5ef3335196f575058fe45f9844e", - "tox_bytes": 51318 + "tox_sha256": "c5fa558601d592cc465e2cf6d1ca6563000049f4936774908ac50edf594c17fe", + "tox_bytes": 51310 } diff --git a/effects/paint/paint.tox b/effects/paint/paint.tox index bb4c9cd..58e7f75 100644 Binary files a/effects/paint/paint.tox and b/effects/paint/paint.tox differ diff --git a/effects/particles/particles.json b/effects/particles/particles.json index 5ca688e..f730670 100644 --- a/effects/particles/particles.json +++ b/effects/particles/particles.json @@ -1,12 +1,12 @@ { "version": "0.4.0", - "sources_sha256": "0b7827608a446f5b7bd188ad2135a6057d55a95bd7e14ea06c0f647a2a258896", + "sources_sha256": "dcf56ca0174552f2ad48ad135412cd088b20d6a4ef91fab292faaa12760d3251", "inputs": [ "in1_top", "in2_audio", "in3_top" ], "touchdesigner_build": "2025.33230", - "tox_sha256": "c2abacfb3a52b0498d73d78572e0faf1a4bc1089d97f4e9c5cc1f0b10d3ccc68", + "tox_sha256": "b802ba91fa0b33e26f6daa514531734eba8d2d346d2491c634cdd2af5ac4a77b", "tox_bytes": 70590 } diff --git a/effects/particles/particles.tox b/effects/particles/particles.tox index 3f71a80..21604b4 100644 Binary files a/effects/particles/particles.tox and b/effects/particles/particles.tox differ diff --git a/effects/pencil/pencil.json b/effects/pencil/pencil.json index 15a4988..f2ed3cd 100644 --- a/effects/pencil/pencil.json +++ b/effects/pencil/pencil.json @@ -1,12 +1,12 @@ { "version": "0.1.0", - "sources_sha256": "9bac13d44d74435f8bf3eb55919a91d28903caa1ce377ec9bc2ccc7933ed78ce", + "sources_sha256": "3b4c050d772c66affaf69324e0f2aa3535bc4c69b4fd482262765b1307d865e0", "inputs": [ "in1_top", "in2_audio", "in3_top" ], "touchdesigner_build": "2025.33230", - "tox_sha256": "bb3bc3b0c37074510ec1a0db739b538528c63c3860538c6353a3e4afe481281e", + "tox_sha256": "8a56bd9261704153d85edf1cb30fec108a1b2496330e47b4febc8cb83eae753b", "tox_bytes": 63126 } diff --git a/effects/pencil/pencil.tox b/effects/pencil/pencil.tox index 87e79f4..5e3e189 100644 Binary files a/effects/pencil/pencil.tox and b/effects/pencil/pencil.tox differ diff --git a/effects/pixel_art/pixel_art.json b/effects/pixel_art/pixel_art.json index c0ccbf4..4a7817b 100644 --- a/effects/pixel_art/pixel_art.json +++ b/effects/pixel_art/pixel_art.json @@ -1,12 +1,12 @@ { "version": "0.1.0", - "sources_sha256": "2a92418dbcb18f7eaf1df07f15abf6477ebd07c0ef4276f5f6d0cdc02cfcfadf", + "sources_sha256": "e635a50c8f653ff12b03df928a1f9ab2f2b57395f37b114eaa330f57d32785b7", "inputs": [ "in1_top", "in2_audio", "in3_top" ], "touchdesigner_build": "2025.33230", - "tox_sha256": "33168e9ae605f14e2eb4c3b6fe8b936fd8a61511c9a01f4772f41a3eb173f0b8", + "tox_sha256": "6b88e6150105dead868e9dfb8f39d4aa5c28801085a85e6159483a764cecb7b0", "tox_bytes": 62502 } diff --git a/effects/pixel_art/pixel_art.tox b/effects/pixel_art/pixel_art.tox index a2dead2..623e2ea 100644 Binary files a/effects/pixel_art/pixel_art.tox and b/effects/pixel_art/pixel_art.tox differ diff --git a/effects/pixel_sort/pixel_sort.json b/effects/pixel_sort/pixel_sort.json index 3f996b9..6d50da3 100644 --- a/effects/pixel_sort/pixel_sort.json +++ b/effects/pixel_sort/pixel_sort.json @@ -1,12 +1,12 @@ { "version": "0.1.1", - "sources_sha256": "ddbdb493ccd9d0347655a6d6cde9779c7f22b59281a87e4898b17a759dff51e2", + "sources_sha256": "a3d32c644d01e1fb7966af66c569c031200f4423499e8c7b1210c8938ed74ed8", "inputs": [ "in1_top", "in2_audio", "in3_top" ], "touchdesigner_build": "2025.33230", - "tox_sha256": "50d3845fb9d7b8c9f72470af1bdcafede16c1c8a6ef5f0db7f08f52b1f4ca4de", + "tox_sha256": "0dea72623967614b64d064100b900ed9d5c1f2b84000f17dde8a28e40195d5ff", "tox_bytes": 38854 } diff --git a/effects/pixel_sort/pixel_sort.tox b/effects/pixel_sort/pixel_sort.tox index d608e68..c33328c 100644 Binary files a/effects/pixel_sort/pixel_sort.tox and b/effects/pixel_sort/pixel_sort.tox differ diff --git a/effects/point_cloud/point_cloud.json b/effects/point_cloud/point_cloud.json index 4db28e6..94e013d 100644 --- a/effects/point_cloud/point_cloud.json +++ b/effects/point_cloud/point_cloud.json @@ -1,12 +1,12 @@ { "version": "0.1.4", - "sources_sha256": "64f1a23d4f59113ab25e536eb30cbd795570b09930b19c370730629e5fa99681", + "sources_sha256": "eff1d3bcd24afbe909cb39beb661cece624fe906cc6515550fe5a3d5d94e5768", "inputs": [ "in1_top", "in2_audio", "in3_top" ], "touchdesigner_build": "2025.33230", - "tox_sha256": "f152ec7d0670ffb22690aaa40b7f81b27aa2b5138de49aa7cf20b66a49edca5a", + "tox_sha256": "55873bf27c8e61f66823bd3dfb5e2a1fa34a8a4022739bb25613628fc74ac0d9", "tox_bytes": 46542 } diff --git a/effects/point_cloud/point_cloud.tox b/effects/point_cloud/point_cloud.tox index 6a74e08..cd2ffa1 100644 Binary files a/effects/point_cloud/point_cloud.tox and b/effects/point_cloud/point_cloud.tox differ diff --git a/effects/reaction_diffusion/reaction_diffusion.json b/effects/reaction_diffusion/reaction_diffusion.json index d4f24b8..d4bcf08 100644 --- a/effects/reaction_diffusion/reaction_diffusion.json +++ b/effects/reaction_diffusion/reaction_diffusion.json @@ -1,12 +1,12 @@ { "version": "0.1.0", - "sources_sha256": "bf554e8ce33388ab73f51710cfba7758bb3093a8f1ac16e281eb8213080c71af", + "sources_sha256": "1f11e339858b6fd4e1277e8a1f9c1f594d4502b64431f68b23b8e2519741355b", "inputs": [ "in1_top", "in2_audio", "in3_top" ], "touchdesigner_build": "2025.33230", - "tox_sha256": "f33f37525dd1a5189879563b1afb213bd07bb74794e23d5b4a5f3886c56abd5d", + "tox_sha256": "2eb1edebc60193dfcc1df3fd9538506b4462039fc7c454e4a989ae39dd711b50", "tox_bytes": 48446 } diff --git a/effects/reaction_diffusion/reaction_diffusion.tox b/effects/reaction_diffusion/reaction_diffusion.tox index 5b03a6e..8f2c127 100644 Binary files a/effects/reaction_diffusion/reaction_diffusion.tox and b/effects/reaction_diffusion/reaction_diffusion.tox differ diff --git a/effects/relight/relight.json b/effects/relight/relight.json index 95fc222..9b90438 100644 --- a/effects/relight/relight.json +++ b/effects/relight/relight.json @@ -1,12 +1,12 @@ { "version": "0.1.0", - "sources_sha256": "9a379aa43e654ed35d0a23aa36faee1fb503132ec54b2c5b057bb6bd4dd77e4f", + "sources_sha256": "84cfd38a3de0b0ea3622950151d4967bfff7c4ee1751bc771ccaa91d850b916c", "inputs": [ "in1_top", "in2_audio", "in3_top" ], "touchdesigner_build": "2025.33230", - "tox_sha256": "499c4c373315d3a6224dd30247ae2751b4fac5158c9aa3b0986087844de34039", + "tox_sha256": "d701cf632971d931646854a4b2c403e72daca453248f66d8674b18b2da1d306f", "tox_bytes": 42294 } diff --git a/effects/relight/relight.tox b/effects/relight/relight.tox index e2386e7..05368c1 100644 Binary files a/effects/relight/relight.tox and b/effects/relight/relight.tox differ diff --git a/effects/scan_processor/scan_processor.json b/effects/scan_processor/scan_processor.json index 91db46d..3a5cfe6 100644 --- a/effects/scan_processor/scan_processor.json +++ b/effects/scan_processor/scan_processor.json @@ -1,11 +1,11 @@ { "version": "0.1.0", - "sources_sha256": "ae2c73a3886f9d0e471939cd0a41d4479abb689cbb4f76d5331aa267d8f5b175", + "sources_sha256": "70ac06303b1f5d873e91ee737df17f09ddcbdbe676f963b58152c1e69d2adc2e", "inputs": [ "in1_top", "in2_audio" ], "touchdesigner_build": "2025.33230", - "tox_sha256": "2c82d8bba962051f12fd31a369bf0f77e439fc4e1b4b7d5dacba12d31eced5d1", + "tox_sha256": "62f0c70ba957b1ff2a4a199559389dabc4499a99f09ba302c1a28a887f103712", "tox_bytes": 29934 } diff --git a/effects/scan_processor/scan_processor.tox b/effects/scan_processor/scan_processor.tox index 7c9642d..7131c5c 100644 Binary files a/effects/scan_processor/scan_processor.tox and b/effects/scan_processor/scan_processor.tox differ diff --git a/effects/slit_scan/slit_scan.json b/effects/slit_scan/slit_scan.json index 7f738f2..40b0035 100644 --- a/effects/slit_scan/slit_scan.json +++ b/effects/slit_scan/slit_scan.json @@ -1,12 +1,12 @@ { "version": "0.1.1", - "sources_sha256": "036684ecdb47616a8ddce52a695577e0e851d3b8b32e1837804d4e7eff9162c9", + "sources_sha256": "67625add0703f0a671c759943394548c21bc8880d2f2537381e7a822759d312f", "inputs": [ "in1_top", "in2_audio", "in3_top" ], "touchdesigner_build": "2025.33230", - "tox_sha256": "e933057bdde9dc560d8e6038266e89a89f2856e5e73ecb760f7501ede87f0300", - "tox_bytes": 54638 + "tox_sha256": "0c98addf816a35bbc8ff6d5764962c140e71af764ec84a8ee72a59d33b9b9864", + "tox_bytes": 54630 } diff --git a/effects/slit_scan/slit_scan.tox b/effects/slit_scan/slit_scan.tox index 8535675..9452808 100644 Binary files a/effects/slit_scan/slit_scan.tox and b/effects/slit_scan/slit_scan.tox differ diff --git a/effects/temporal_resonator/temporal_resonator.json b/effects/temporal_resonator/temporal_resonator.json index 877703d..ecbd48a 100644 --- a/effects/temporal_resonator/temporal_resonator.json +++ b/effects/temporal_resonator/temporal_resonator.json @@ -1,12 +1,12 @@ { "version": "0.1.1", - "sources_sha256": "263125dc4b32e47db265a586e3a7e11148e10495def27a7800494e6a32417624", + "sources_sha256": "35ddfe93c0394a3ec9c973ff8b29ba6745add76e5241b19c3769394d43d0797e", "inputs": [ "in1_top", "in2_audio", "in3_top" ], "touchdesigner_build": "2025.33230", - "tox_sha256": "67ef0634916acdb9f29d76717ae15f698ccc09579ca2da116ba56dffe840e98f", + "tox_sha256": "f1789c87acaecb26f24a51dda157ddc2829e700541e70ca1cd29b51aabff75f8", "tox_bytes": 40390 } diff --git a/effects/temporal_resonator/temporal_resonator.tox b/effects/temporal_resonator/temporal_resonator.tox index eb37faf..c893a53 100644 Binary files a/effects/temporal_resonator/temporal_resonator.tox and b/effects/temporal_resonator/temporal_resonator.tox differ diff --git a/effects/vhs_crt/vhs_crt.json b/effects/vhs_crt/vhs_crt.json index 23ed805..b2acd4d 100644 --- a/effects/vhs_crt/vhs_crt.json +++ b/effects/vhs_crt/vhs_crt.json @@ -1,11 +1,11 @@ { "version": "0.2.2", - "sources_sha256": "9656f93d55fb5e2dea9694b785881a910b401390bc6d2e79641ef071ad9f1ad0", + "sources_sha256": "dda8471e3b1ca8da5b3b343b37ecc5800e2bef013ebb9cbc8b711a5d84481ef8", "inputs": [ "in1_top", "in2_audio" ], "touchdesigner_build": "2025.33230", - "tox_sha256": "5e49ec255ed696274f0673077196c12e4772eda18589a82c0885c6e2d27a70f7", + "tox_sha256": "bc99c4240e3fb32d180abe424431f0d964a90ba2524fe5afef82b74e13763849", "tox_bytes": 78358 } diff --git a/effects/vhs_crt/vhs_crt.tox b/effects/vhs_crt/vhs_crt.tox index 5e37634..dd5792d 100644 Binary files a/effects/vhs_crt/vhs_crt.tox and b/effects/vhs_crt/vhs_crt.tox differ diff --git a/effects/video_vocoder/video_vocoder.json b/effects/video_vocoder/video_vocoder.json index 0a982b7..3713030 100644 --- a/effects/video_vocoder/video_vocoder.json +++ b/effects/video_vocoder/video_vocoder.json @@ -1,12 +1,12 @@ { "version": "0.1.0", - "sources_sha256": "99a64e6c2ab2573bba5338121875b00a09171643b69cb4d1280660641cad3928", + "sources_sha256": "890223bc1233ebf877f6eedae395c0a9bfbf249557de1212439f3f79c32e174f", "inputs": [ "in1_top", "in2_audio", "in3_top" ], "touchdesigner_build": "2025.33230", - "tox_sha256": "6ed9293a0fb8a7ab0aae4b4cdc96a92a1c1f0b81e072a91835c276346ac70316", + "tox_sha256": "7a4620724537d176d9dbeb13da4babd8ac356719bfa23f82cce972e923293ab7", "tox_bytes": 42054 } diff --git a/effects/video_vocoder/video_vocoder.tox b/effects/video_vocoder/video_vocoder.tox index e661ed3..a06d82a 100644 Binary files a/effects/video_vocoder/video_vocoder.tox and b/effects/video_vocoder/video_vocoder.tox differ diff --git a/effects/voxel_space/voxel_space.json b/effects/voxel_space/voxel_space.json index f32680e..c7bdf26 100644 --- a/effects/voxel_space/voxel_space.json +++ b/effects/voxel_space/voxel_space.json @@ -1,12 +1,12 @@ { "version": "0.1.0", - "sources_sha256": "3ce60280d27371b77856a4006794390077e4b4a93f54d3451584ef418421309a", + "sources_sha256": "3c4ba7f03bb87f7fe440228f476ee6f399ef9e84247cece9abd26f07350807e1", "inputs": [ "in1_top", "in2_audio", "in3_top" ], "touchdesigner_build": "2025.33230", - "tox_sha256": "09249a79a4f0b821df9be81cae6625b68493b605ef5331fd6d1d5d3fc6defdce", + "tox_sha256": "1c1ff829b08406067efa172bae39b4e2ea8eb9d55919fd2942c67dfcc571e31e", "tox_bytes": 43766 } diff --git a/effects/voxel_space/voxel_space.tox b/effects/voxel_space/voxel_space.tox index c33a5c1..bdd62cc 100644 Binary files a/effects/voxel_space/voxel_space.tox and b/effects/voxel_space/voxel_space.tox differ diff --git a/effects/words/words.json b/effects/words/words.json index c60e7bf..e83a531 100644 --- a/effects/words/words.json +++ b/effects/words/words.json @@ -1,12 +1,12 @@ { "version": "0.1.0", - "sources_sha256": "5ea2cbeebb65b785396f57d15bc70efdf9452df3336f96fc6c0e69eb92838159", + "sources_sha256": "4eac783d340b5faec55dece99f937dcc102a79aeb8ad08bffb64b0101db362ea", "inputs": [ "in1_top", "in2_audio", "in3_top" ], "touchdesigner_build": "2025.33230", - "tox_sha256": "e54254f20615392d4155cab16417cb7c754f400c4e5f9396123ad18196d30b5c", + "tox_sha256": "21c70c56305c43bf28a3e1db396326d38dd05ebe3111199ade3a0da0847ada04", "tox_bytes": 66446 } diff --git a/effects/words/words.tox b/effects/words/words.tox index df26a2c..bf60ee5 100644 Binary files a/effects/words/words.tox and b/effects/words/words.tox differ diff --git a/effects/fluid/src/divergence.glsl b/lib/glsl/divergence.glsl similarity index 60% rename from effects/fluid/src/divergence.glsl rename to lib/glsl/divergence.glsl index 99d4e5d..3efde86 100644 --- a/effects/fluid/src/divergence.glsl +++ b/lib/glsl/divergence.glsl @@ -1,6 +1,6 @@ -// Fluid (touchdesigner-effects, MIT): how much fluid each cell gains or loses (the divergence of the -// velocity), which the pressure solve then cancels so the fluid neither piles up nor thins out -// (Harris 2004, GPU Gems ch. 38). +// touchdesigner-effects (MIT), shared by the effects that run a fluid (lib/glsl, incompressible()): Fluid. How much +// fluid each cell gains or loses (the divergence of the velocity), which the pressure solve then cancels so the fluid +// neither piles up nor thins out (Harris 2004, GPU Gems ch. 38). // Inputs: 0 the velocity before projection (cells per second). Out: r the divergence (per second). out vec4 fragColor; vec2 vel(vec2 p) { return texture(sTD2DInputs[0], p).rg; } diff --git a/lib/glsl/pressure.glsl b/lib/glsl/pressure.glsl new file mode 100644 index 0000000..8bd61dc --- /dev/null +++ b/lib/glsl/pressure.glsl @@ -0,0 +1,14 @@ +// touchdesigner-effects (MIT), shared by the effects that run a fluid (lib/glsl, incompressible()): Fluid. The +// pressure that cancels the divergence, by Jacobi iteration of its Poisson equation (Harris 2004, GPU Gems ch. 38). +// The GLSL TOP runs this shader Passes times a frame, each pass reading the last one's result as input 0 +// (2025.33230, dev/probes/glsl_passes.py); the first pass starts from last frame's pressure, which is close, so a few +// dozen passes do. Off the frame's edges the pressure reads 0 (Input Extend Mode Zero): the fluid flows out freely. +// Inputs: 0 the pressure so far (last frame's on the first pass), 1 the divergence. Out: r the pressure. +out vec4 fragColor; +float p(vec2 q) { return texture(sTD2DInputs[0], q).r; } +void main() { + vec2 px = uTD2DInfos[0].res.xy; + vec2 uv = vUV.st; + float sum = p(uv + vec2(px.x, 0.0)) + p(uv - vec2(px.x, 0.0)) + p(uv + vec2(0.0, px.y)) + p(uv - vec2(0.0, px.y)); + fragColor = vec4(0.25 * (sum - texture(sTD2DInputs[1], uv).r), 0.0, 0.0, 1.0); +} diff --git a/effects/fluid/src/project.glsl b/lib/glsl/project.glsl similarity index 59% rename from effects/fluid/src/project.glsl rename to lib/glsl/project.glsl index 9c0e2d7..b3f5f8d 100644 --- a/effects/fluid/src/project.glsl +++ b/lib/glsl/project.glsl @@ -1,6 +1,6 @@ -// Fluid (touchdesigner-effects, MIT): the velocity made incompressible: minus the pressure's gradient -// (the projection of Stam 1999; Harris 2004, GPU Gems ch. 38). This is the velocity the dye moves with -// and the next frame starts from. +// touchdesigner-effects (MIT), shared by the effects that run a fluid (lib/glsl, incompressible()): Fluid. The +// velocity made incompressible: minus the pressure's gradient (the projection of Stam 1999; Harris 2004, GPU Gems +// ch. 38). This is the velocity what the fluid carries moves with, and the next frame starts from. // Inputs: 0 the velocity before projection, 1 the pressure. Out: rg the velocity (cells per second). out vec4 fragColor; float p(vec2 q) { return texture(sTD2DInputs[1], q).r; } diff --git a/lib/glsl/spectrum.glsl b/lib/glsl/spectrum.glsl new file mode 100644 index 0000000..9b6c970 --- /dev/null +++ b/lib/glsl/spectrum.glsl @@ -0,0 +1,21 @@ +// touchdesigner-effects (MIT), shared by the effects that sum light over wavelengths (lib/glsl, lib_glsl()): Lens +// Flare's starburst. The CIE 1931 color matching functions by C. Wyman, P.-P. Sloan and P. Shirley's analytic fit +// ("Simple Analytic Approximations to the CIE XYZ Color Matching Functions", JCGT 2(2), 2013), a sum of piecewise +// Gaussian lobes per function, and the matrix from CIE XYZ to linear sRGB (D65). + +float lobe(float lam, float mu, float s1, float s2) { + float t = (lam - mu) / (lam < mu ? s1 : s2); + return exp(-0.5 * t * t); +} + +// x-bar, y-bar, z-bar at a wavelength (nm). +vec3 cmf(float lam) { + return vec3(1.056 * lobe(lam, 599.8, 37.9, 31.0) + 0.362 * lobe(lam, 442.0, 16.0, 26.7) + - 0.065 * lobe(lam, 501.1, 20.4, 26.2), + 0.821 * lobe(lam, 568.8, 46.9, 40.5) + 0.286 * lobe(lam, 530.9, 16.3, 31.1), + 1.217 * lobe(lam, 437.0, 11.8, 36.0) + 0.681 * lobe(lam, 459.0, 26.0, 13.8)); +} + +const mat3 XYZ_TO_RGB = mat3(3.2404542, -0.9692660, 0.0556434, + -1.5371385, 1.8760108, -0.2040259, + -0.4985314, 0.0415560, 1.0572252); diff --git a/lib/glsl/warmth.glsl b/lib/glsl/warmth.glsl new file mode 100644 index 0000000..fa7d0a2 --- /dev/null +++ b/lib/glsl/warmth.glsl @@ -0,0 +1,37 @@ +// touchdesigner-effects (MIT), shared by the effects whose picture warms what it touches (lib/glsl, warmth()): +// Frost thaws by it. One heat field from what stands out in front (or inside the mask) and what moves, each +// spread a little around itself, eased so it rises quickly and falls slowly: a hand that moves warms the glass +// at once and leaves it to cool behind it. A clip slower than the project repeats each frame (change reads 0 +// between its frames), so the movement is held from the last new frame rather than read as stillness. On the first +// frame (the feedback's alpha is 0) the change is against nothing, so nothing has moved yet. A change below a dead +// zone warms nothing: haze drifting through a dark room changes it a little everywhere. +// Inputs: 0 what stands out or the mask (r 0-1; sets the size), 1 the change since the last frame (lib +// change.glsl: r |change|, at 1/8), 2 the clock (lib clock.glsl: r 1 on a new input frame), 3 this pass's last +// output (Feedback TOP). +// Out: r the warmth 0-1, g the movement held from the last new frame. +uniform vec4 uW; // x how much what stands out warms, y how much movement warms, z the change that warms fully, + // w the change below which nothing warms +uniform vec4 uR; // x this frame's step (s), y the time to rise (s), z the time to fall (s) +out vec4 fragColor; + +// A 3 x 3 tent around uv in input i's texels: what warms spreads a little around itself. +float spread(int i, vec2 uv) { + vec2 o = uTD2DInfos[i].res.xy; + float s = 4.0 * texture(sTD2DInputs[i], uv).r; + s += 2.0 * (texture(sTD2DInputs[i], uv + vec2(o.x, 0.0)).r + texture(sTD2DInputs[i], uv - vec2(o.x, 0.0)).r + + texture(sTD2DInputs[i], uv + vec2(0.0, o.y)).r + texture(sTD2DInputs[i], uv - vec2(0.0, o.y)).r); + s += texture(sTD2DInputs[i], uv + o).r + texture(sTD2DInputs[i], uv - o).r + + texture(sTD2DInputs[i], uv + vec2(o.x, -o.y)).r + texture(sTD2DInputs[i], uv + vec2(-o.x, o.y)).r; + return s / 16.0; +} + +void main() { + vec2 uv = vUV.st; + vec4 last = texture(sTD2DInputs[3], uv); + bool fresh = texelFetch(sTD2DInputs[2], ivec2(0), 0).r > 0.5; + float moved = last.a == 0.0 ? 0.0 : fresh ? smoothstep(uW.w, uW.z, spread(1, uv)) : last.g; + float target = max(uW.x * spread(0, uv), uW.y * moved); + float t = target > last.r ? uR.y : uR.z; + float w = mix(target, last.r, exp(-uR.x / max(t, 1e-3))); + fragColor = vec4(w, moved, 0.0, 1.0); +} diff --git a/lib/tox_build.py b/lib/tox_build.py index 1302a8e..eb36cd1 100644 --- a/lib/tox_build.py +++ b/lib/tox_build.py @@ -320,6 +320,45 @@ def flow_speed(comp, root, mo, start, dead=0.25, hold=0.3): return pace, motion +def incompressible(comp, root, advected, passes=40, carries='the dye'): + """The velocity made incompressible, shared by the effects that run a fluid (Fluid): divergence (how much fluid each + cell gains or loses) -> pressure (+ pressure_fb: Jacobi iterations of its Poisson equation, passes of them in one + TOP, starting from last frame's) -> velocity (advected minus the pressure's gradient), the projection of stable + fluids (Stam 1999) as Harris laid it out for the GPU (GPU Gems ch. 38, 2004). advected is the velocity before + projection (rg cells per second; it sets the size), carries what the fluid carries, for the comments. Every pass + is 32-bit float; the divergence holds at the frame's edges, and the pressure reads 0 off them, so the fluid flows + out freely. The shaders are lib/glsl's divergence, pressure and project. Returns (divergence, pressure, + velocity).""" + divergence = glsl(comp, 'divergence', [advected], lib_glsl(root, 'divergence.glsl'), + 'How much fluid each cell gains or loses.', fmt='rgba32float') + pressure_fb = mk(comp, feedbackTOP, 'pressure_fb', [divergence], 'The pressure last frame.', top='pressure') + pressure = glsl(comp, 'pressure', [pressure_fb, divergence], lib_glsl(root, 'pressure.glsl'), + 'The pressure that cancels the divergence: {} Jacobi iterations.'.format(passes), fmt='rgba32float') + setpars(pressure, npasses=passes) + velocity = glsl(comp, 'velocity', [advected, pressure], lib_glsl(root, 'project.glsl'), + 'The velocity made incompressible: what {} moves with.'.format(carries), fmt='rgba32float') + setpars(divergence, inputextenduv='hold') + return divergence, pressure, velocity + + +def warmth(comp, root, subject, mo, start, subject_warms=1.0, move_warms=1.0, full=0.04, dead=0.01, rise=0.15, + fall=1.5): + """Warmth, shared by the effects whose picture warms what it touches (Frost thaws by it): what stands out in + front or inside the mask (subject: r 0-1, which sets the size) times subject_warms, and what moves (optical_flow()'s + ops, mo: change and clock; a change of full in luma a frame warms fully, one below dead not at all) times + move_warms, each spread over 3 x 3 texels, as one + field that rises over rise seconds and falls over fall seconds. Movement is held from the last new input frame, + so a clip slower than the project doesn't read as still between its frames. start is a 1 x 1 TOP, black with alpha + 0, the Feedback TOP (warmth_fb) starts from. Any of the weights and times may be an Expr. The shader is lib/glsl's + warmth. Returns the warmth TOP (r the warmth 0-1).""" + fb = mk(comp, feedbackTOP, 'warmth_fb', [start], "warmth's last frame.", top='warmth') + w = glsl(comp, 'warmth', [subject, mo['change'], mo['clock'], fb], lib_glsl(root, 'warmth.glsl'), + 'What the picture warms: what stands out or is inside the mask, and what moves; quick to rise, slow to ' + 'fall.', fmt='rgba16float', + uW=(subject_warms, move_warms, full, dead), uR=(Expr('min(absTime.stepSeconds, 0.1)'), rise, fall)) + return w + + def flash_gate(comp, trigger, name='flash'): """Every flash an effect makes goes through here (docs/TD_IDIOMS.md A1). A Trigger CHOP turns trigger's channels into one-frame-wide pulses no closer together than the Flashing parameter