open-ai-canvas brings AI filmmaking onto a node graph

AI short-film production often fragments across script documents, reference folders, generation tools, and download directories. open-ai-canvas, also known as 无限画布 or Infinite Canvas, consolidates those materials into a self-hosted, node-based production workbench.

Built with a TypeScript front end and Go back end, the repository has 81 GitHub stars. It began as a fork of infinite-canvas and adds a multi-user backend, structured storyboard pipeline, asset management, agent integrations, and a 3D director view.

Production lives on the graph

The canvas connects scripts, character cards, style references, scenes, shot lists, generated media, and director-stage outputs. Each node records part of the production, giving creators reusable character and scene references across multiple shots.

• Canvas tools: multiple projects, connected nodes, box selection, automatic layout, undo and redo, a minimap, import and export, and public read-only sharing.
• Media generation: text, image, video, and audio tasks, along with reference-image editing, first- and last-frame controls, camera motion, video continuation, and local inpainting.
• Film workflow: structured storyboard scripts, character cards, batch shot nodes, and a 3D stage for composing scenes and writing control images back to the canvas.
• Task management: asynchronous queues, logs, failure retries, asset storage, and account-based backend synchronization.