Infinite Canvas is an open-source, AI-native infinite canvas workbench for visual creation, combining AI image generation, reference-image editing, video generation, and agent assistance on one zoomable board. Published on GitHub under the MIT license, it has earned more than 7,800 stars and is actively developed, with updates in October 2026. A hosted instance runs at canvas.best.
Infinite Canvas is designed as a workspace for AI creation rather than a single-purpose generator. On the canvas you can arrange generated images, reference pictures, and video clips, then keep iterating visually instead of juggling files and chat threads. It integrates AI image generation, reference-image editing, video generation, an agent assistant, canvas orchestration, conversational creation, a prompt library, and material management, and it supports multi-agent collaborative workflows. The app is compatible with the OpenAI API ecosystem and can connect to a variety of relay and gateway channels, so users can plug in their preferred model providers. It is built with Vite and can be deployed in one click to services like Render, and its prompt library and asset management features target creators who run many variations before settling on a final piece.
Infinite Canvas is free and open source under the MIT license. Self-hosting costs only your infrastructure, and the canvas.best instance lets you try it in the browser. You bring your own model API access, so generation costs depend on the providers and channels you connect.
Infinite Canvas is for AI creators, designers, and content teams who think spatially: storyboard-style image exploration, campaign asset iteration, and video pre-visualization. If single-prompt generator pages feel cramped and you want a persistent board where an entire project lives, this workbench fits.