Text-to-Image
Diffusers
Safetensors
StableDiffusionPipeline
stable-diffusion
stable-diffusion-diffusers
Instructions to use CompVis/stable-diffusion-v1-4 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use CompVis/stable-diffusion-v1-4 with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("CompVis/stable-diffusion-v1-4", dtype=torch.bfloat16, device_map="cuda") prompt = "A high tech solarpunk utopia in the Amazon rainforest" image = pipe(prompt).images[0] - Notebooks
- Google Colab
- Kaggle
- Local Apps
- Draw Things
- DiffusionBee
Copying noise / Scaling noise / space filling noise
#52
by xalex - opened
To allow for changing the image size and still getting similar images, it would be an interesting idea to copy the initialization from a smaller image to a larger when re-rendering to get similar results in the area where the noise was copied. One could also try to scale the noise, but I wonder if this would result in blocky images.
Another idea for brainstorming would be to distribute the noise using a space filling curve, starting with the noise from the smaller image and then adding new noise afterward.