Turns a library of green-screen cat clips into narrated POV story reels (vertical, 1080×1920) in the style of cat-meme shorts. You describe each clip once; then you write a story in plain words and desired emotions, and the engine matches clips, drops them onto scene-relevant backgrounds, labels them, and renders the video.
# 1. (one time) describe every clip as text -> data/catalog.json
python3 engine/build_catalog.py
# 2. render a story -> output/final.mp4
python3 engine/render.py data/stories/functional-adult.jsonSee what a feeling resolves to: python3 engine/match.py sleepy relaxed.
engine/ code (build_catalog · match · render · paths)
clips/ source green-screen clips (git-ignored, re-downloadable)
backgrounds/ bundled AI scene library (preferred over the web)
data/ catalog.json + stories/*.json (committed)
work/ regenerable artifacts (git-ignored)
output/ finished reels (git-ignored)
docs/ full documentation ← start at docs/README.md
Playlist → clips → described once as text (catalog.json) → you write a story in
words → a matcher turns each desired emotion into a clip → ffmpeg keys the cats onto
scene backgrounds, grounds them, adds the POV bubble / labels / captions, and stitches
the beats into a reel.
Full docs: docs/ — architecture, the renderer's grounding math,
the background system, the design study behind the style, story-authoring guide, and the
decisions/gotchas.
