OpenKatana: An Open-Source AI Video Editor That Lets Claude Drive ComfyUI
rk0neki · reddit · 2026-10-29
- What it is: OpenKatana, an entry in Comfy's Open Source Higgsfield challenge — an open-source local video editor driven by an AI agent, Apache-2.0 licensed, no account required, no forced cloud services.
- Architecture: Provides 85 nodes and 6 Apps inside ComfyUI; the Full workflow combines project ID, asset upload, Make edit→Render, 3D starts, and 72 editing tools in a single graph. ComfyUI, Studio, CLI and MCP all share the same project file, so humans and agents can collaboratively edit the same video.
- Demo: The author drags in anime clips, music, and stickers, writes a one-line brief for Claude, and Claude handles the edit via Studio, the render engine, scripts, and skills. The intro and outro of the launch video were both cut with OpenKatana.
- Positioning: Ships with no generative models — the agent is user-provided (Claude), and generative models are installed separately. Apps run locally only and refuse paid/cloud jobs. Windows/Linux CI supported; macOS untested; Studio is not a full 3D editor.
- Demo video: YouTube, code: GitHub.
More from coding & agent
- Dev shows GPT-6 Astra staging a JUCE recreation of the Roland TB-303 synth — kylegawley · 2026-10-29
- Espressif launches ESP-Mosaico: an ESP32-S31 kit built for vibe coding — songguoxiansen · 2026-10-29
- Dev shares design notes for a remote MCP server letting 20+ Claude projects share one board — mmmmjjjj1228 · 2026-10-29
- Dev post-trains Qwen3-4B into a tool-using agent: 66 MB llama.cpp LoRA runs on a Mac, val 0.51→0.83 — DesignerIsland5524 · 2026-10-29
- Dev builds read-only Grok Bot assistant that answers questions from your Mac screen and call history — bfrench · 2026-10-29
- Dev recalls 5,000 GitHub contributions in one day: "I could see reasoning traces in my sleep" — vincent_koc · 2026-10-29