Vedit
Edit videos via AI instructions or an interactive timeline interface powered by a local, deterministic FFmpeg engine.
Vedit combines a traditional non-linear video editing interface with agentic AI capabilities, exposing over 100 tools over the Model Context Protocol (MCP). Instead of generating closed, uneditable video clips, it manipulates a real timeline—allowing AI agents and human editors to execute the exact same operations such as trimming, splitting, pacing to music, and color grading. It runs completely locally on your hardware with no subscriptions or cloud uploads, keeping all project files structured as deterministic JSON and enabling seamless integration with tools like Claude Code, Cursor, and Codex.
主な機能
01Dual interface supporting both a browser-based timeline UI and comprehensive MCP server tools
02Deterministic rendering engine backed locally by FFmpeg with GPU acceleration
03Agent visual feedback tools including frame rendering and contact-sheet grid previews
04Smart audio and timeline analysis including BPM detection, beat-aligned cuts, and speech silence removal
05Full project state stored in editable JSON for version control, automated diffs, and scripting
0617 GitHub stars
ユースケース
01Delegating rough cuts, silence removal, and b-roll placement to an AI coding agent like Claude Code or Cursor
02Automating social media clip generation and reformatting (e.g., 16:9 to 9:16) for YouTube Shorts and TikTok
03Scripting programmatically driven video pipelines and batch edits using structured JSON project definitions