SankaiAI/TwitCanva-Video-Workflow ? reverse-engineered prompt

Build me a modern AI canvas app where I can drag and connect nodes to create images and videos from text, reference images, or other generated assets.

I want a clean workspace with an infinite canvas, smart type checked connections, an asset library, and the ability to save and reload workflows. It should let me generate images and videos with different AI providers, switch models based on what I’m feeding in, and include things like image to image, text to image, text to video, frame to frame, storyboard creation, camera angle edits, and motion transfer from a reference video. Add a built in chat assistant that can help guide the workflow too.

Please also make it easy to post finished images or videos to X and TikTok if the user connects their accounts. Use a secure backend so API keys never get exposed in the browser, and make it work nicely as a React and TypeScript app with a fast local dev setup. If you need to check current docs for any of the AI services, look them up online first.