An infinite canvas for AI-generated images, video, and 3D scenes
It exists to keep AI image generation, video rendering, and 3D scene blocking in one continuous workspace instead of scattered browser tabs.

What it does
infinite-canvas is an open-source workbench that treats AI media generation like a whiteboard session. You drag nodes around an infinite canvas to generate images, edit references, render video, and even block out 3D scenes with a “director’s console” that manages cameras and environments. A built-in agent can analyze your selected nodes, plan the next creative step, and drop the result back onto the board. It speaks standard OpenAI-compatible APIs and supports a laundry list of third-party image and video models.
The interesting bit
The unusual part is the “director’s console”: a 3D viewport where you place models, set panoramic backgrounds, and position cameras, then automatically feed those camera frames back into the canvas as image nodes for further AI processing. It is essentially a tiny virtual production stage inside a node editor. The camera settings—focal length, aperture, and lens type—even get baked into the generation prompts automatically.
Key highlights
- Agent-driven generation that reads upstream canvas nodes and writes results back to the board
- 3D director’s console with camera management, panoramic environments, and screenshot-to-node export
- OpenAI-compatible API support for images, video, and chat, plus third-party models like Seedance and Kling
- Workflow templates with variables, reference inputs, and reusable prompt/asset libraries
- Go backend with SQLite and S3-compatible storage, replacing the original pure-frontend architecture
Caveats
- Explicitly flagged as early-stage development with no guarantee of backward compatibility for stored data
- Database schemas and storage formats are subject to direct, breaking changes without migration paths
- The authors recommend forking if you need a stable, independently maintained branch
Verdict
Worth a look if you want a self-hostable, visual control room for chaining AI image and video generation. Skip it if you need production-grade data stability or a simple, single-purpose tool.
Frequently asked
- What is tigerowo/infinite-canvas?
- It exists to keep AI image generation, video rendering, and 3D scene blocking in one continuous workspace instead of scattered browser tabs.
- Is infinite-canvas open source?
- Yes — tigerowo/infinite-canvas is open source, released under the AGPL-3.0 license.
- What language is infinite-canvas written in?
- tigerowo/infinite-canvas is primarily written in TypeScript.
- How popular is infinite-canvas?
- tigerowo/infinite-canvas has 1k stars on GitHub.
- Where can I find infinite-canvas?
- tigerowo/infinite-canvas is on GitHub at https://github.com/tigerowo/infinite-canvas.