28998306/magicalcanvas

MagicalCanvas is a Windows desktop application that combines AI content generation with video editing, aimed at Chinese-speaking users who want to create short videos or illustrated content from text without using a command line or browser. The interface is entirely in Chinese and the app requires no installation of external tools. The first half of the app is a node-based canvas where you connect building blocks in a visual flow. Each block can be a piece of text, an image, or a video clip. You draw lines between blocks to define the sequence: for example, text flows into an image generation step, which flows into a video generation step. You can feed a full script into the canvas and have the AI automatically break it into character, scene, and shot nodes, generate all the images, and arrange them. The canvas supports text-to-image, image-to-video, and text-to-video generation. The second half is a built-in video editing studio with a multi-track timeline. Tracks handle video, voiceover audio, subtitles, and stickers separately. Within the editor you can adjust clip speed, add transitions from a built-in library, set brightness and contrast per clip, layer stickers and screen effects, and export finished video at 720p or 1080p. FFmpeg, the tool that handles actual video processing, is bundled inside the app so there is nothing extra to install. The app connects to AI models through any OpenAI-compatible API, meaning you supply your own API keys and endpoint addresses in settings. Four categories of model can be configured independently: text generation, image generation, video generation, and speech recognition for automatic subtitle creation. For voice narration, the app includes Microsoft Edge text-to-speech, which is free and supports multiple Chinese voice styles. All API keys and generated assets stay on your local machine and are not sent to any third party. The project is built on Electron and React, is Windows-only, and is licensed under Apache 2.0.