Figma meets Stable Diffusion, but make it AGPL

What it does

Infinite Canvas is a browser-based workspace for visual AI workflows. You drag nodes around an infinite board, generate images via OpenAI-compatible APIs, edit reference pictures, manage prompts, and chat with an assistant that can see your selected canvas nodes. Everything lives in one React/Next.js frontend backed by a Go service.

The interesting bit

The “canvas assistant” is the unusual part: it reads the upstream context of whatever nodes you’ve selected, generates images or answers from that context, and drops the results back onto the board. It’s less a drawing tool and more a visual REPL for iterative image generation.

Key highlights

• OpenAI-compatible API support, including chatgpt2api, grok2api, flow2api, and newapi adapters
• Video generation via Seedance 2.0 through Volcano Ark Agent Plan
• Curated prompt library scraped from open-source GitHub projects
• Docker Compose deployment with Render one-click deploy
• AGPL-3.0 licensed

Caveats

• Explicitly flagged as pre-release: database schemas and storage formats may change without migration paths
• Authors advise against public multi-user deployments; intended for personal or local use
• No topics tagged on GitHub, limited discoverability

Verdict

Worth a spin if you’re iterating heavily on AI-generated visuals and want a self-hosted alternative to closed platforms like Midjourney’s web UI. Skip it if you need stability, team collaboration, or a mature plugin ecosystem.