basketikun/infinite-canvas

What it solves

Infinite Canvas is an open-source workbench designed for image creation. It solves the problem of fragmented workflows by consolidating canvas orchestration, AI image generation, reference image editing, a conversational assistant, and prompt libraries into a single, unified interface. This allows creators to explore visual schemes and iteratively refine image results in one place.

How it works

The project provides an infinite canvas where users can drag, scale, and connect nodes. It connects directly to OpenAI-compatible APIs via the browser to perform text-to-image, image-to-image, reference image editing, text-based Q&A, and audio/video generation. It includes a "Canvas Assistant" that can interact with selected nodes and their upstream nodes to generate content and insert results back into the canvas.

Who it’s for

Visual artists, AI image creators, and designers who need a non-linear, node-based workspace for iterative AI-driven visual exploration and content generation.

Highlights

• Infinite Canvas: Supports multi-canvas projects, node dragging, scaling, and connections, featuring a mini-map and undo/redo functionality.
• AI Integration: Direct browser-to-API connection for multimodal generation (images, video, audio) and text interaction.
• Local Agent Support: Integrates with Codex/Claude Code via a local Canvas Agent and MCP (Model Context Protocol) to allow agents to operate the canvas.
• Extensible Plugin System: Supports dynamic installation and updating of remote node plugins via URL, with a TypeScript SDK for custom development.
• Prompt Library: Built-in access to seven open-source prompt sources with local caching via IndexedDB.
• Customizable API Calls: Flexible configuration for image and video generation scripts to adapt to various API providers.