wtf is ai-image-gen?

ai-image-gen is a small, local tool for generating images and videos using the SiliconFlow API. You bring your own SiliconFlow API key, pick a model, type a prompt, and it makes the image or video for you. There are no accounts to create and no backend server, everything runs on your own machine in the browser, and your API key never leaves it. The project is aimed at people who want a local, no-login, easily modified interface rather than SiliconFlow's own web playground, which is fine for a few quick one-off generations but not built for tinkering or forking. It is built with Vue 3 and Vite, and you run it with Bun. To get started you copy a provided .env.example file to .env and either paste your API key into the app once it is running, which saves it in your browser local storage, or set it directly in the .env file so it is already filled in when you open the app. Because browsers cannot call the SiliconFlow API directly due to missing CORS headers, the app uses Vite development proxy to forward your requests along with your key. SiliconFlow runs two separate platforms, an international one and a China based one, each with its own accounts and keys, so a key issued for one will be rejected by the other. The tool offers two modes you can switch between. Image mode works right away and lets you choose from models such as Kolors, FLUX.1, Qwen-Image, and SD 3.5 Large, along with settings like size, batch count, steps, guidance, and seed. Video mode is slower: it submits a job to generate a short video from text or from a source image, then checks back until the render is ready and plays it once finished. Both modes let you paste in any other SiliconFlow model id instead of the built-in choices. The project is released under the MIT license.