Desktop AI art engine that combines a node graph, specialized editors and a 3D director stage for generating short-video, game and advertising assets, with local-first model access.
262+6 this week38TypeScriptOtherWell maintainedUpdated 10 hours ago
raph651/open-source-canvas
An open-source node canvas for generating images and video by wiring prompts, references and models into workflows run on a local ComfyUI server.
- Stars
- 143
- Forks
- 37
- Open issues
- 0
- Last commit
- 7 days ago
What is it?
Open Source Canvas is a node-based studio for image and video generation. Prompts, reference images and models are placed as nodes on an infinite board and connected with lines, and finished images and clips come out as outputs. The front end is built with React, Vite and React Flow, and it does no generation itself: all work is sent to a ComfyUI server over its HTTP and WebSocket API.
Local models that run out of the box are SD-Turbo for images and LTX-Video 2B for video. Hosted models such as LTX-2.5 and Seedream 5.0 Pro are listed through ComfyUI partner nodes, but generating them directly from the canvas is not yet wired and fails with a clear message. Boards are saved as plain JSON in the browser, with autosave, version history and export.
The project includes a studio home with example and template boards, per-node generation controls for prompt, references, model, resolution and seed, image-to-image and image-to-video chaining, and an optional WebSocket collaboration room. A production server mode proxies ComfyUI and keeps each visitor's outputs separate.
Key features
- 01Infinite node canvas with typed connections, undo/redo, grouping and copy/paste.
- 02Per-node generation panel for prompt, references, model, resolution, aspect, duration and seed.
- 03Image-to-image and image-to-video chains where an upstream image feeds the next node.
- 04Runs against a local ComfyUI server with SD-Turbo and LTX-Video 2B as default models.
- 05Boards are stored as plain JSON with local autosave, version history and export.
- 06Optional collaboration room that syncs a board over WebSocket.
Use cases
- Prototyping image or short video sequences from prompts on your own machine.
- Chaining image generation into image-to-video for storyboards or motion tests.
- Running a shared ComfyUI instance for several people with separated outputs.
Quick start
- 1
Clone the repo and install app dependencies
git clone https://github.com/raph651/open-source-canvas.git && cd open-source-canvas && npm install - 2
Set up ComfyUI and download the default models (or follow the manual steps in the README)
./scripts/setup-comfyui.sh --with-video - 3
Start the app in development mode, then open http://127.0.0.1:4173
npm run dev
Things to know
- Needs a local ComfyUI server running alongside the app.
- LTX-Video 2B needs about 24 GB of unified or GPU memory at default settings.
- API models such as Seedream and LTX-2.5 cannot yet generate from the canvas and need a custom wrapper.
- Some model weights have restrictive licenses, such as LTX-2.5 for large companies and MiniMax H3's community license.
In collections
Topics
FAQ
- What is open-source-canvas?
- An open-source node canvas for generating images and video by wiring prompts, references and models into workflows run on a local ComfyUI server. Open Source Canvas is a node-based studio for image and video generation. Prompts, reference images and models are placed as nodes on an infinite board and connected with lines, and finished images and clips come out as outputs. The front end is built with React, Vite and React Flow, and it does no generation itself: all work is sent to a ComfyUI server over its HTTP and WebSocket API.
- Is open-source-canvas free to use? What license does it have?
- open-source-canvas is open source under the MIT license. Check the license terms before using it in commercial products.
- Is open-source-canvas still maintained?
- Actively maintained. The last commit to open-source-canvas was on Nov 1, 2026.
- What can I use open-source-canvas for?
- Prototyping image or short video sequences from prompts on your own machine. Chaining image generation into image-to-video for storyboards or motion tests. Running a shared ComfyUI instance for several people with separated outputs.
- How do I install open-source-canvas?
- Clone the repo and install app dependencies: git clone https://github.com/raph651/open-source-canvas.git && cd open-source-canvas && npm install Set up ComfyUI and download the default models (or follow the manual steps in the README): ./scripts/setup-comfyui.sh --with-video Start the app in development mode, then open http://127.0.0.1:4173: npm run dev
- Where can I get open-source-canvas?
- The source code and installation instructions are on GitHub: https://github.com/raph651/open-source-canvas
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
A collection of custom nodes and quality-of-life improvements for ComfyUI, the node-based Stable Diffusion interface.
3.7K+15 this week280JavaScriptMITWell maintainedUpdated 9 days ago
A node-graph GUI, API and backend for running diffusion models locally to generate images, video, audio and 3D content.
AI & Machine Learning#stable-diffusion#diffusion-models#node-editor#image-generation#video-generation
145.6K+1.2K this week17.3KPythonGPL-3.0Updated 7 hours ago
A modular web UI for AI image, video and audio generation, built on Stable Diffusion, Flux, Wan and other models, with a ComfyUI workflow tab.
4.8K+22 this week483C#MITWell maintainedUpdated yesterday
A Gradio-based web UI for running Stable Diffusion and newer image/video models, continuing the Forge fork with added model support and optimizations.
1.9K+23 this week264PythonAGPL-3.0Updated 2 days ago
The official web front-end for ComfyUI, the node-based interface for running generative AI image and video workflows.
2.1K+9 this week764TypeScriptGPL-3.0Updated 2 hours ago