GitHub Trending has surfaced browser-use's video-use project, which describes itself as 100% open source: users drop raw footage into a folder and chat with a coding agent that has shell access, such as Claude Code or Codex, to get a final.mp4 back. According to the project README, the pipeline removes filler words and dead space, auto color grades each segment, adds 30ms audio fades at every cut, burns in subtitles and generates animation overlays through parallel sub-agents. Its core approach packs ElevenLabs Scribe word-level transcripts into a roughly 12KB text file as the model's main reading surface, rendering timeline filmstrips only at decision points instead of feeding frames to the model.