Instructions to use stabilityai/stable-diffusion-xl-base-0.9 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use stabilityai/stable-diffusion-xl-base-0.9 with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("stabilityai/stable-diffusion-xl-base-0.9", dtype=torch.bfloat16, device_map="cuda") prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k" image = pipe(prompt).images[0] - Notebooks
- Google Colab
- Kaggle
- Local Apps
- Draw Things
- DiffusionBee
Doesn't work well with DreamFusion, generates bad quality images under high guidance
#51
by JunzheJosephZhu - opened
Seems that under high CFG the generated image is more noisy.
Above is 20 step DDIM from SDXL, under guidance=100, resolution=512x512, conditioned on resolution=1024, target_size=1024
Below is 20 step DDIM from SD2.1 under guidance=100, resolution=512x512, conditioned on resolution=1024, target_size=1024
Why would you use such a high guidance scale? That will ruin the image no matter what AI you're using. The fact that the bad images look slightly different than you're used to doesn't make them any better or worse than the previous bad images. SDXL is a different setup than SD, so it seems expected to me that things will behave a little differently.

