Following the leak of SDXL 0.9 model, the official 1.0 is finally released two days ago. The Automatic1111’s web ui is also updated for you to use the new models. Read on to find out how to update webui and use SDXL 1.0.
- I assume that you already have a working installation of Automatic1111’s Web UI. Please install it before you follow the steps.
- Update your webui to at least 1.5.1 by issuing this command at your installation directory:
- Download the base model from this page and put it under models/Stable-diffusion:
- (Optional) Download the lora model from this page and put it under models/Lora:
- Download the refiner model from this page and put it under models/Stable-diffusion:
- Restart webui if you have it running
- You might have to modify your COMMANDLINE_ARGS. I added these two arguments:
- Click on txt2img tab. Select the sd_xl_base model and make sure VAE set to Automatic and clip skip to 1.
- Use a prompt of your choice. Set the size to width to 1024 and height to 1024. Click on GENERATE to generate an image.
- Click on Send to img2img button to send this picture to img2img tab.
- Change the checkpoint to the refiner model.
- Use a low denoising strength, I used 0.15.
- Remove any Lora from your prompt if you have them. Click on GENERATE to generate the image.
Some points to note:
- Don’t use Lora for previous SD versions. TI from previous versions are Ok.
- Size cheat sheet.
- A model titled sd_xl_base_1.0_0.9vae.safetensors is also included in the repo. You can try to download it here. Its corresponding refiner model is here.
- It’s not very stable, you might have to restart webui often.
This post may contain affiliated links. When you click on the link and purchase a product, we receive a small commision to keep us running. Thanks.