Instructions to use stabilityai/stable-diffusion-3-medium-diffusers with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use stabilityai/stable-diffusion-3-medium-diffusers 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-3-medium-diffusers", dtype=torch.bfloat16, device_map="cuda") prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k" image = pipe(prompt).images[0] - Inference
- Notebooks
- Google Colab
- Kaggle
- Local Apps
- Draw Things
- DiffusionBee
Why negative prompt need extra conditioning?
#31
by RyanStark - opened
And when I ignore these conditioning nodes, link the negative prompt to ksampler directly, nothing become worse or better. So these extra nodes really confused me, maybe someone can explain it.
You're asking a comfyui related question in the diffusers model, probably will be better to ask comfyui questions in the appropriate repo. Anyways, those nodes just mean that the negative prompt gets applied in the 0.1 (10%) of the generation and for the rest they get zeroed. SD3 doesn't really use that much the negative prompt so that's why you practically don't see any difference.
