ComfyArt: a local Lovart-style infinite canvas on your ComfyUI
tldraw + React infinite canvas as a ComfyUI custom node. No account, no credits, no cloud. Optional design agent over MCP. Image plus short video via bundled MiniMax H3 workflows.
ComfyArt by @jtydhr88 (MIT, created Oct 13, about 20 stars when we checked) is a fully local Lovart-style infinite canvas on your own ComfyUI. The UI is tldraw plus React. You install it as a custom_nodes package and open http://127.0.0.1:8188/comfyart/. There is no account, no credit meter, and no cloud dependency for the canvas itself. Docs live at https://comfyart.org/en. Note from the README: tldraw ships under its own license in dist, separate from the MIT project licence.
Beyond drawing on the board, it can drive image and short video work. Bundled workflows cover MiniMax H3 image-to-video and text-to-video. Canvas tools need the ComfyArt tab open while you use them — a practical gotcha if you expect everything to run from the default Comfy graph view alone.
Optional design agent: wire it through MCP to Claude Code, Codex, Qwen Code, DeepSeek Harness, or a local LLM, so an agent can help on the same local canvas. That is the piece that makes it interesting for ReelStack readers who already run agent-plus-ComfyUI stacks and want a design surface without a SaaS canvas bill.
Our take: early (days old, small star count), but the local-first canvas + MCP agent combo is rare. We haven't installed it. Credit @jtydhr88; start from the announce thread and the README install path.
Related tools on ReelStack
Skills & repos in our directory
- ComfyArt · ComfyUI, Image & design
- FastH3 Omni · ComfyUI, Generation
- seriesmith · ComfyUI, Generation
- ReAngle · ComfyUI, VFX & compositing
- Comfield · Repo, Generation
- FXjutsu · ComfyUI, VFX & compositing
More from the Lab
- VoiceStudio: local voice cloning, dubbing and an MCP server for your video agentSpotted, not yet tested
- #OpenSourceHiggsfield: Comfy's $25K challenge and the first open rebuildsSpotted, not yet tested
- MiniMax H3 gets fast locally: Sol-H3 Studio, HyperFlow 8-step, Ref2VA-VSASpotted, not yet tested
- FastH3 Omni goes open: reference-to-video with sound on your own GPU, and the case for letting Opus cut cheap clipsSpotted, not yet tested
Topics: ComfyUI · Open-weight video & audio models · Video MCP servers