feat(deck): real robot photographs replace all three placeholder slates (deck-real-photos) - #29
Conversation
|
/agentic_review |
PR Summary by QodoReplace presentation deck placeholders with real robot photos
AI Description
Diagram
High-Level Assessment
Files changed (11)
|
The drop-in contract held exactly as designed: three real photographs converted to WebP at the manifest's existing pixel targets and saved over the same filenames, so no `src`, `thumbSrc`, or dimension changed. Slot assignment follows what each frame actually shows, not the source filenames: - reachy-mini-hero <- a Reachy Mini table portrait facing the lens, antennas rising out of frame, blurred marble fireplace behind (cropped 4:3 -> 16:9, 1920x1080) - reachy-mini-action <- a Reachy Mini close-up mid-gesture, head tilted up and back on its jointed neck (already ~3:2, 1800x1200) - so101-hero <- the SO-ARM101 raised and reaching in side profile, gripper open (cropped 4:3 -> 16:9, 1920x1080) Alt text and `subject` are rewritten per slot. This is not cosmetic: the previous text described *intended* shots and is now factually wrong about these frames — it claimed "a single camera eye" (Reachy has two lenses), "a person's hand beside it for scale" (no hand is in the action frame), and the arm "at rest ... folded into the home pose" (it is extended and reaching). A screen-reader user is told what the alt says, so it has to match the photograph that ships. Module and doc prose likewise drop the placeholder framing. A fourth candidate (a desk scene with a laptop and hands) went unused: at 1003x564 it is below the action slot's 1800x1200 target and would be soft projected. Verified: 20/20 node tests, astro build clean, 23/23 presentation checks, 56 pytest, markdownlint clean; slide 4, the close cards, and the index thumbs screenshot-reviewed. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01FqTq59W6wi1YZqxGf228Tt
809c744 to
edf1bc4
Compare
|



Real robot photographs replace all three placeholder slates. The drop-in
contract held exactly as designed — the photos were converted to WebP at the
manifest's existing pixel targets and saved over the same filenames, so no
src,thumbSrc, or dimension changed.Slot assignment
By what each frame actually shows, not by the source filenames:
reachy-mini-heroreachy-mini-actionso101-heroA fourth candidate (a desk scene with a laptop and a person's hands) went
unused: at 1003×564 it is well below the action slot's 1800×1200 target and
would look soft projected.
Alt text rewritten — not cosmetic
The previous alt text described intended shots and is factually wrong about
these frames. It claimed "a single camera eye" (Reachy has two lenses),
"a person's hand beside it for scale" (no hand is in the action frame), and
the arm "at rest … folded into the home pose" (it is extended and
reaching). A screen-reader user is told what the alt says, so it has to match
the photograph that ships. Each slot's
subjectand the shot-brief doc areupdated the same way, and both drop the placeholder framing.
Verification
20/20 node tests, astro build clean, 23/23 presentation checks, 56 pytest,
markdownlint clean. Slide 4, the close cards (dark theme), and the Presentations
index thumbs screenshot-reviewed. File sizes stay in the documented band:
39–98 KB full, 6–11 KB thumbs. Version 0.18.0 → 0.18.1.
🤖 Generated with Claude Code
https://claude.ai/code/session_01FqTq59W6wi1YZqxGf228Tt