How mengtang1114-forks/video-use works
video-use is an agent-guided video editing workflow: the user supplies footage and approves an edit strategy; an agent uses transcripts and visual drill-downs to create an edit decision list (EDL), renders and evaluates the result, then returns a final video. The README documents the workflow and capabilities; sampled helper code supports transcription, transcript packing, timeline views, and rendering. The orchestration and several documented features are described but not sampled, so those links reflect documentation rather than verified implementation calls.
Main entry points: render.py Video rendering · transcribe.py Transcription · SKILL.md Cut evaluation · timeline_view.py Visual drill-down
Components 22
Editing workflow
- Editing rules
SKILL.md - Inventory and strategy
SKILL.md - Strategy approval
SKILL.md - Edit decision list
- Project memory
- Cut evaluation
SKILL.md - Final video
Speech and review
- Transcription
helpers/transcribe.py - Batch transcription
helpers/transcribe_batch.py - Word transcripts
- Transcript packing
helpers/pack_transcripts.py - Packed transcript
- Visual drill-down
helpers/timeline_view.py
Render and polish
- Video rendering
helpers/render.py - Color grading
helpers/grade.py - Animation overlays
skills/manim-video
External integrations
- ElevenLabs Scribe
- FFmpeg tools
- Overlay engines
Other components
- Editor
- Editing agent
- Source footage
Connections 32
- Editor → Source footage: supplies
- Editor → Editing agent: requests edit
- Editing agent → Editing rules: follows
- Editing agent → Inventory and strategy: inspects
- Source footage → Inventory and strategy: inventoried by
- Editing agent → Strategy approval: proposes strategy
- Editor → Strategy approval: approves
- Source footage → Batch transcription: transcribed from
- Source footage → Transcription: transcribed from
- Batch transcription → Transcription: dispatches
- Transcription → FFmpeg tools: extracts audio
- Transcription → ElevenLabs Scribe: uploads audio
- Transcription → Word transcripts: writes
- Word transcripts → Transcript packing: reads
- Transcript packing → Packed transcript: writes
- Editing agent → Packed transcript: reads
- Editing agent → Visual drill-down: requests view
- Source footage → Visual drill-down: frames and audio
- Word transcripts → Visual drill-down: labels view
- Editing agent → Edit decision list: authors
- Edit decision list → Video rendering: drives render
- Video rendering → FFmpeg tools: invokes
- Video rendering → Color grading: applies grading
- Video rendering → Animation overlays: composites
- Animation overlays → Overlay engines: uses
- Video rendering → Final video: produces
- Final video → Cut evaluation: inspected at cuts
- Cut evaluation → Visual drill-down: requests views
- Editing agent → Cut evaluation: reviews output
- Cut evaluation → Edit decision list: issues prompt fixes
- Editing agent → Project memory: persists session
- Final video → Editor: delivers