0xsline/OpenChatCut

Open-source, local-first conversational AI video editor with a professional multi-track timeline, Agent Skills, MCP integration, and Remotion rendering.

What is OpenChatCut?

OpenChatCut is an open-source video editor that puts AI agents (like the built-in chat assistant, or external tools like Codex and Claude Code) inside a real, professional editing timeline. Instead of one-shot AI video generation that produces a final, un-editable clip, every AI edit lands on actual tracks, clips, transitions, captions, and effects in the project file — you can pause, undo, adjust, or hand it to another agent.

It's a local-first desktop/web app built with TypeScript, React, Electron, and the Remotion rendering library. It's not affiliated with the commercial "ChatCut" product and is released under an AGPL license.

Why it's interesting

Real timeline + natural language: You can edit by hand like in Premiere or Final Cut, or describe a change and let an agent do it — both work on the same project.
Agent-native: Uses the Model Context Protocol (MCP), so external coding agents can read, edit, and export your video projects directly.
Rich editing features: Multitrack video/audio, WebGL effects, LUTs, transitions, keyframes, motion graphics, even in-browser person segmentation (so captions avoid the speaker's face).
Transcript-driven editing: Automatically transcribe video/audio, then cut out pauses or mistakes by editing the text.
Local-first: Your projects and media stay on your machine by default; AI API keys stay server-side.
Production-ready exports: MP4, audio, captions, FCPXML (Final Cut Pro), and full project data.

How it works (core loop)

Describe the goal → Agent reads the project → Proposes edits → Writes to timeline → Preview/undo → Manage captions & audio → Export.

Key capabilities at a glance

Area	What you get
Timeline	Multitrack editing, trim, ripple, keyframes, undo/redo
Visuals	WebGL effects, LUTs, chroma key, custom shaders
Audio	Multi-track, vocal isolation, auto‑ducking, voice‑over
Transcript	Word‑level transcription, pause compression, speakers
Captions	Auto‑generate, styled, translatable, SRT export
Motion Graphics	Templates + agent‑generated editable clips
AI generation	Images, video, speech, music, sound effects
Media	Uploads, online search, folder management
Export	MP4, audio, captions, FCPXML, hardware accelerated
Publishing	Agent‑driven posting to social platforms (with preview & confirmation)
Agent	Built‑in chat agent + external MCP connections

Getting started

Installers: Prebuilt DMG (macOS), Windows x64, and Linux AppImage from GitHub Releases.
From source: Requires Node.js 24.x. npm install then npm run dev and open http://localhost:5199.
The editor needs a secure context — use localhost or HTTPS.
The built‑in agent can work with API keys (OpenAI, Anthropic) or via the official Codex CLI for ChatGPT subscribers. Claude Code works through a local MCP connection.

Status

Actively developed, early builds. The project format and agent tools will keep evolving.

Note: This explanation is based solely on the project's README. Check the repository for the latest updates.