video-use: Delegating Video Editing to Coding Agents

Place the original video in a folder, and a coding agent like Claude Code performs the editing on your behalf. It handles everything from removing unnecessary speech or silence, automatic color correction, and subtitle insertion to generating animated overlays, all through conversation.

The LLM does not view the video directly; instead, it reads ElevenLabs transcription data and, when necessary, generated visual composite images. This enables precise cut editing using 12KB of text and a few images, rather than the 45 million tokens required to analyze 30,000 frames.

After rendering, it self-evaluates each cut boundary to detect visual jumps or audio pops. Only results that pass are shown to the user, and project memory is saved to allow editing to continue in the next session.