Skip to content

README.md

Latest commit

 

History

History
71 lines (55 loc) · 2.18 KB

README.md

File metadata and controls

71 lines (55 loc) · 2.18 KB
Raw

Engine examples

Real media (the primary example)

The fullest example is testdata/engine/real/: a real KAS:ST track ("Astral Talk", 174 BPM) and five real, annotated dive clips, converted from an actual prior DiveSync session. See testdata/engine/real/README.md for provenance and the grid-derivation arithmetic.

Media itself is not committed (~250 MB; see media/README.md). With the media bank present at media/:

.venv/bin/python -m divesync.engine render \
  --project testdata/engine/real/project.json \
  --out output/sample \
  --profile peak_time \
  --seed 7

Produces a real, playable, vertical (1080x1920) skeleton edit at output/sample/skeleton.mp4, plus every intermediate artefact (timeline.json, state.json, targets.json, placement.json, rhythm.json, effects.json, planner-comparison.json, decision-log.jsonl, summary.json).

Without media (annotation-only / CI-safe)

Every stage from annotation loading through planning works without any video or audio file on disk — durations are simply left None and logged. To exercise the full pipeline down to (but not including) the FFmpeg render:

.venv/bin/python -m divesync.engine render \
  --project testdata/engine/real/project.json \
  --out output/no_video \
  --no-video

Style profiles

The same inputs under the three shipped profiles produce measurably different targets (docs/configuration.md has the deltas table):

for profile in hypnotic peak_time broken; do
  .venv/bin/python -m divesync.engine render \
    --project testdata/engine/real/project.json \
    --out "output/sample-$profile" --profile "$profile" --no-video
done

Grid diagnostics

.venv/bin/python -m divesync.engine grid \
  --music testdata/engine/real/astral-talk.music.json \
  --duration 30 --click /tmp/click.wav

Prints the first beat/bar/phrase positions, the anchor-residual table, a Cm/Ca confidence trace, and writes an audible click track to /tmp/click.wav for a by-ear grid-accuracy check (spec 26).

Validation only

.venv/bin/python -m divesync.engine validate \
  --music testdata/engine/real/astral-talk.music.json \
  --clip testdata/engine/real/dive-04.clip.json