DiffSynth Studio v1.0
The first version of DiffSynth Studio has been released!
In this project, we provide a new diffusion engine. This diffusion engine is very efficient. We compared it with other implementations.
Resolution: 1024*1024, model architecture: Stable Diffusion XL, steps: 30
| Diffusion engine | Time consumed for generating an image |
|---|---|
| DiffSynth Studio | 5.0s |
| Stable Diffusion Webui | 6.8s |
The images generated by DiffSynth Studio are a little cleaner. Here are some examples:
| DiffSynth Studio | Stable Diffusion Webui |
|---|---|
 |
 |
We also provide a simple UI, which supports real-time painting with SD-XL-Turbo. Enjoy it!
This project is still in development. We will add more interesting features in the future.