Video Editing
Inspect the recording and transcript, then choose complete thought units. Read editing workflows. Keep the original recording and capture source-relative cut boundaries in the project.
The footage workflow executes explicit segments in listed order. Optional removeSilence intersects those segments with detected audible ranges. Silence detection is an audio threshold heuristic; review its output around breaths, laughter, emphasis, and quiet speakers. It does not identify filler words or choose the best take semantically.
Use source-relative transcript words when retaining the original audio; the runner remaps them through every cut. When replacing the audio with narration, provide words on the narration timeline or transcribe the new narration.
Default to contain to preserve a code demo or face. cover is a center crop and does not track subjects. The built-in footage layout reserves space for captions but cannot guarantee face safety. Inspect the rendered crop. Editable cut decisions are retained as edit-decisions.json.
Professional timelines, beat analysis, automatic visual take scoring, and application MCP integrations are reference guidance in this release.
For fast reels, use reel direction. Cut on meaning or action, vary shot scale, and preserve a self-contained claim and its proof. A rapid rhythm does not justify removing a qualifier that changes the speaker’s meaning.