tdsoc2002/pincanvas

PinCanvas is a visual, node-based canvas for building AI-powered creative workflows. You connect nodes together on a canvas, where each node can generate an image, a video, or text using AI models of your choice. The connections between nodes pass material from one step to the next, so you can chain together complex sequences without writing any code. The tool supports a range of AI providers, including OpenAI-compatible services, Midjourney, and several Chinese platforms such as Seedance and Jimeng. You can configure which services to use through a settings panel in the browser, entering your own API keys and base URLs. No account registration is required for the app itself. All project data is stored locally in your browser using a built-in storage mechanism, so nothing leaves your machine unless you explicitly enable cloud storage for uploaded files. Node types cover common creative tasks: generating images from text, generating video from images or text, combining image and audio inputs, and using reference images by mentioning them directly in a prompt. A typical workflow might take a single text description, pass it to an image generation node, and then feed that image into a video generation node. Each node can be configured independently for model, resolution, and duration. The project is built with React 18 and TypeScript and uses a visual flow library called xyflow to handle the canvas and connections. State is managed with Zustand, and the build tooling runs on Bun. An optional backend handles object storage if you want generated files saved to an S3-compatible service, but this is not required for basic use. Adding new AI models is supported through the settings UI without touching any code, which makes it practical for people who want to try different providers without modifying the source. For those who do want to extend the code, the README describes exactly which files to change and in what order.