Video Editing
AI-assisted editing for real footage — not generation from prompts. Core thesis: the value is not generation, it is compression.
Workflow
Applies §10.1 KISS (deterministic FFmpeg cuts over reinvented pipelines) and §10.2 YAGNI (compress real footage; generate assets only where footage is missing).
- Load contexts: read
.ai-engineering/manifest.ymlproviders.stacks; load.ai-engineering/overrides/<stack>/conventions.mdper stack +.ai-engineering/overrides/_shared/conventions.md; load.ai-engineering/team/*.md. - Gate — verify
ffmpeg(ffmpeg -version); install viabrew install ffmpeg/apt install ffmpeg/choco install ffmpeg. - Pick mode —
plan(design the edit structure from raw footage/transcript),organize(transcribe, label, identify segments, generate edit decision lists/EDL),cut(deterministic FFmpeg trim/split/concat/reframe/normalize),compose(programmable Remotion overlays + compositions, optional). - Run the 6-layer pipeline — Capture → Organization → Deterministic Cuts → Programmable Composition → Generated Assets → Final Polish (human).
- Cross-reference
ai-mediafor Layer 5 generated assets (voiceover, music/SFX, b-roll).
Detail: 6-layer pipeline + tool table, FFmpeg recipes: extract / batch-cut / concat / proxy / silence detect, social-platform reframing presets.
Common Mistakes
- Forcing one tool to span every layer.
- Ignoring proxy / audio-normalization hygiene.
Integration
Called by: user directly, /ai-build. Calls: ffmpeg (deterministic cuts), Remotion (compositions), /ai-media (Layer 5 generated assets). See also: /ai-media (asset generation), /ai-slides (deck embeds), /ai-visual (cover art).
Examples
User: "cut this 60-minute talk into a 90-second highlight reel"
/ai-video-editing plan recording.mp4
Plans cuts, transcribes, identifies highlight beats, runs FFmpeg trim+concat, normalizes audio, outputs the reel.
$ARGUMENTS