rain0993/videoediting

video-use is a tool that lets you edit raw video footage by working with an AI coding agent such as Claude Code, Codex, or similar tools that can run shell commands. Instead of using a traditional video editor with menus and presets, you drop your raw footage into a folder, describe in plain language what you want the final video to look like, and the agent produces a finished file named final.mp4. The tool handles several editing tasks automatically: it cuts out filler words like umm and uh along with dead space between takes, applies color grading to each clip, adds short audio fades at cut points to avoid pops, and burns in subtitles in a default two word uppercase style that can be customized. It can also generate animated overlays using several external tools, run in parallel so multiple animations are created at once. Under the hood, the agent does not watch the video directly. Instead, it reads a text transcript with word level timestamps and speaker labels generated by the ElevenLabs Scribe transcription service, which keeps the amount of data the AI needs to process small compared to processing every video frame. For decisions that need visual context, such as ambiguous pauses or comparing retakes, it generates a still image combining a filmstrip, waveform, and word labels for that section instead of full video frames. Setup requires cloning the repository, installing ffmpeg, and providing an ElevenLabs API key for transcription. The project is designed to be installed as a skill inside a compatible AI agent, and it keeps a memory file so a later editing session can continue where a previous one left off. Before producing a final video, the tool automatically checks its own output at each cut point and will fix and re render up to a few times if it finds problems like visual jumps or audio glitches. The README does not state a license.