video-use

About: video-use is an open-source conversational video editing skill from the Browser Use team. It turns shell-capable coding agents such as Claude Code, Codex, Hermes, and Openclaw into a local editing assistant. The user drops raw footage into any folder, starts the agent there, describes the target edit, reviews the proposed strategy, and receives outputs under the local edit/ directory, including preview.mp4 or final.mp4.

Core capabilities: first, it uses ElevenLabs Scribe for word-level timestamps, then packs multi-take transcripts into takes_packed.md so the LLM can reason over text before touching cuts. Second, timeline_view generates filmstrip and waveform views only at decision points, keeping visual review targeted. Third, edit decisions live in edl.json and render.py extracts segments, concatenates them, applies overlays, and burns subtitles at the end of the filter chain. Fourth, grade.py wraps FFmpeg grading, while the included SKILL.md documents hard production rules such as word-boundary cuts, short audio fades, output-timeline SRT offsets, and subtitle-last rendering.

Best for: talking-head clips, launch videos, tutorials, interviews, multi-take montages, travel footage, and subtitle-heavy social videos. It is not a hosted SaaS editor; it is a local, auditable toolchain that lets an agent operate video editing scripts.

Pricing: MIT-licensed and free to self-host. Real usage requires local Python dependencies and FFmpeg; transcription requires bringing an ElevenLabs API key.

Onboarding: the README provides a setup prompt, while install.md covers cloning the repo, installing dependencies, registering the skill, and configuring the API key. The repository was created in April 2026, had about 15.7k stars on 2026-07-07, and was last pushed on 2026-07-01.