A browser-based implementation of Stable Diffusion using onnxruntime that runs entirely client-side using WebGPU acceleration. Now with an llm chat!

GPU is mandatory unfortunately. Might be Nvidia only. (I think cpu users can still use llm)

Features:

1) Live generation preview

2) 5 different scheduler implementations

3) Generation history

4) 3 supported models (converted existing models to onnx format)

5) A floating llm chat window (for funzies)