What problem does it solve? Editing raw footage into a finished video normally requires manual timeline work in a GUI editor. This Skill lets an AI agent perform the entire edit — transcription, take selection, cutting, grading, animation overlays, and subtitles — through conversation and a typed Node.js CLI, without the LLM ever watching the video frames. ## Core Features & Use Cases - Word-level transcription and packing: ElevenLabs Scribe transcribes footage with speaker diarization, then packs it into a token-efficient phrase-level markdown (takes_packed.md) the agent reads to pick cuts. - EDL-driven render pipeline: The agent writes edl.json with cut decisions, and the CLI renders per-segment extracts with grades, 30ms audio fades, PTS-shifted overlays, and subtitles burned in last. - Self-evaluation and animations: Timeline filmstrip/waveform PNGs verify cut boundaries on the rendered output, and parallel sub-agents build overlay animations with HyperFrames, Remotion, Manim, or PIL. - Use Case: Drop raw interview takes into a folder, ask the agent to "edit these into a launch video", confirm the proposed strategy, and receive a graded, subtitled final.mp4 in <videos_dir>/edit/. ## Quick Start Ask your agent to set up the video-use repository, then say "edit the videos in this folder into a short launch video" and confirm the proposed strategy.
