Pinhole

Tiny text-to-image that runs entirely in your browser. No server, no API key.
Downloads once, then cached by your browser.
Load a model to start.

Pinhole models are 13M to 71M parameters and trained in the Darkroom project on 18,000 images made by PixArt-LCM. They are best at single, simple subjects in three styles: pixel art, flat icon and sticker. Everything runs on your device with ONNX Runtime Web. Models: S · M · L · License: CreativeML Open RAIL++-M.