generate-footage

The plugin's other entry point. video-use edits footage you HAVE. This skill builds videos from footage you DON'T have yet: narration plus still keyframes, composed and animated by HyperFrames, rendered deterministically.

Pipeline (proven in production; see docs/rca/ for why the motion step is deterministic):

script (with v3 audio tags)
  → eleven_v3 narration            (scripts/tts_v3.py)
  → transcribe                     (npx hyperframes transcribe narration.mp3)
  → lock scene windows to sentence boundaries (word-level timestamps)
  → STILLS FROM ANYWHERE           (one keyframe per scene — see below)
  → animate-stills composition     (recipes/04-animate-stills.md, build_scenes.py)
  → lint → render → motion + audio QA (scripts/motion_check.py, contact sheet)