OSS
video-use Edits Video by Giving the Agent a Transcript, Not Frames — 18,972 Stars and MIT-Licensed
The browser-use team's video-use lets you drop raw footage in a folder, chat with Claude Code or another coding agent, and get a finished final.mp4. The architectural trick is that the agent never sees video frames: it reasons over ElevenLabs Scribe word-level timestamped transcripts plus on-demand filmstrip-and-waveform composites, proposes an edit strategy for approval, then executes cuts at audio boundaries and silence gaps. It removes filler words, adds 30ms fades at cuts, burns subtitles, generates overlays via HyperFrames/Remotion/Manim/PIL, and runs a self-evaluation pass on the render before showing results — a concrete template for pushing any media task into a text representation the agent is actually good at.
↳ Follow the thread