What agents does video-use work with?

Any coding agent with shell access, including Claude Code, Codex, and similar tools. It's installed as a skill symlinked into the agent's skills directory.

Does video-use require an internet connection or third-party API?

Yes. Transcription (word-level timestamps, speaker diarization, audio events) uses the ElevenLabs Scribe API, so you need an ElevenLabs API key and network access.

What license is video-use released under?

MIT, so you can use, modify, and redistribute it freely.

How does it avoid feeding raw video frames to the LLM?

It transcribes audio into a packed text file with word-level timestamps and generates on-demand visual composites (filmstrip, waveform, word labels) only at decision points, instead of dumping every frame.

What tools does it use for animation overlays?

It can generate overlays via HyperFrames, Remotion, Manim, or PIL, running each animation in a parallel sub-agent.

Does it require ffmpeg?

Yes, ffmpeg is a required dependency for rendering and color grading; yt-dlp is optional for pulling in online source footage.

video-use: Edit raw footage into a finished

video-use is a skill that turns a coding agent (Claude Code, Codex, or any agent with shell access) into a video editor. Drop raw footage into a folder, describe the edit you want in chat, and the agent produces a final.mp4. It's built for people who'd rather type an editing brief than click through a timeline, and it works on talking heads, tutorials, montages, interviews, and travel footage without presets or templates.

Instead of feeding video frames to an LLM (a token-expensive approach that also just doesn't work well), video-use transcribes audio with word-level timestamps and speaker diarization, packs everything into a small text file, and lets the model reason over that. A visual composite (filmstrip + waveform + word labels) is generated only on demand, at ambiguous decision points, not for every frame.

Key features

Filler and dead-space removal: strips umm, uh, false starts, and silence between takes automatically
Auto color grading: applies warm cinematic, neutral punch, or custom ffmpeg chains per segment
Click-free audio cuts: 30ms audio fades at every cut point so edits don't pop
Subtitle burning: 2-word uppercase chunks by default, fully customizable styling
Parallel animation overlays: spawns sub-agents to generate overlays via HyperFrames, Remotion, Manim, or PIL, one agent per animation
Self-evaluation loop: renders, then checks its own output at every cut boundary for visual jumps, audio pops, or hidden subtitles before showing you a preview (retries up to 3 times)
Session memory: writes a project.md file so a later session can pick up where the last one left off
Text-first pipeline: reads a ~12KB packed transcript instead of scanning thousands of frames, keeping token use low

Ideal use cases

video-use fits workflows where you already have raw footage and want an agent to handle the cutting, pacing, and cleanup: turning a pile of takes into a launch video, tightening a talking-head recording, assembling a tutorial from screen-capture clips, or producing a rough cut of an interview. It's also useful for recurring editing jobs, since session memory lets you resume a project across days.

It's not a fit if you need frame-accurate manual control over every cut, a GUI-based NLE experience, or complex multi-track compositing beyond what ffmpeg-based scripts and the supported animation tools (HyperFrames, Remotion, Manim, PIL) can produce. It also depends on ElevenLabs for transcription, so it's not suited for fully offline or air-gapped setups, and it needs an agent with real shell access (Claude Code, Codex, etc.), not a chat-only interface.

Installation

The simplest path is to paste a setup prompt into an agent with shell access (Claude Code, Codex, or similar):

Set up https://github.com/browser-use/video-use for me.

Read install.md first to install this repo, wire up ffmpeg, register the skill with whichever agent you're running under, and set up the ElevenLabs API key — ask me to paste it when you need it. Then read SKILL.md for daily usage, and always read helpers/ because that's where the editing scripts live. After install, don't transcribe anything on your own — just tell me it's ready and wait for me to drop footage into a folder.

The agent clones the repo, installs dependencies, registers the skill, and prompts once for an ElevenLabs API key (get one at elevenlabs.io/app/settings/api-keys).

To install by hand:

# 1. Clone and symlink into your agent's skills directory
git clone https://github.com/browser-use/video-use ~/Developer/video-use
ln -sfn ~/Developer/video-use ~/.claude/skills/video-use        # Claude Code
# ln -sfn ~/Developer/video-use ~/.codex/skills/video-use       # Codex

# 2. Install deps
cd ~/Developer/video-use
uv sync                         # or: pip install -e .
brew install ffmpeg             # required
brew install yt-dlp             # optional, for downloading online sources

# 3. Add your ElevenLabs API key
cp .env.example .env
$EDITOR .env                    # ELEVENLABS_API_KEY=...

Once installed, point the agent at a folder of raw takes, start a session (claude, codex, etc.), and describe the edit you want. Output lands in <videos_dir>/edit/final.mp4.