video-use, from the Browser Use team, turns a coding agent into a video editor. Released as 100% open source under the MIT license and already past 16,000 GitHub stars, it lets you drop raw footage into a folder, chat with an agent like Claude Code, and get a finished `final.mp4` back — no timeline, no presets, no menus. It reframes editing as a conversation with an agent that drives ffmpeg and a library of editing scripts on your behalf. ## Editing by Conversation The workflow is deliberately minimal: point your agent at a directory of raw takes, say something like "edit these into a launch video," and video-use inventories the sources, proposes a strategy, waits for your approval, then renders the result next to your footage. Because it works through natural language rather than a GUI, it adapts to any kind of content — talking heads, montages, tutorials, travel, interviews — without you learning an editor's interface or wiring up presets. ## Automated Editing Primitives Under the hood video-use bundles the tedious parts of editing as automatic operations. It cuts filler words like "umm" and "uh" along with false starts and dead space between takes, auto color-grades each segment with customizable ffmpeg chains, and applies short 30ms audio fades at every cut so you never hear a pop. It also burns in styled subtitles — two-word uppercase chunks by default — that you can fully customize, turning raw clips into a polished cut. ## Parallel Sub-Agents and Self-Evaluation More ambitiously, video-use can generate animation overlays through tools like HyperFrames, Remotion, Manim, or PIL, spawning parallel sub-agents with one per animation so they render concurrently. It self-evaluates the rendered output at every cut boundary before presenting anything, and persists session state in a `project.md` file so a later session picks up where you left off. This agent-orchestration approach is what lets a single prompt drive a genuinely multi-step editing pipeline. ## Trade-offs and Limitations video-use is a skill you run through a coding agent, so it requires an agent with shell access plus ffmpeg, and some features lean on external services — an ElevenLabs API key for audio and third-party tools for animation. As a young, fast-moving project the results depend on prompt quality and the agent's judgment, and automated cuts should be reviewed rather than trusted blindly for high-stakes edits. It suits creators comfortable in a terminal more than those wanting a polished consumer app. ## Who Should Use This video-use is a natural fit for developers, creators, and technical content producers who already live in Claude Code or a similar agent and want to automate repetitive editing — cutting filler, grading, subtitling, and assembling a first cut. If you value a keyboard-and-prompt workflow over a timeline UI and want an open, hackable, MIT-licensed editing pipeline, video-use is a compelling agent-native option.