Instructions to use stabilityai/stable-diffusion-x4-upscaler with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use stabilityai/stable-diffusion-x4-upscaler 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-x4-upscaler", 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
Is there any ComfyUI implementation for this?
Just wondering if anyone has implemented this in ComfyUI?
It looks like https://github.com/Zuellni/ComfyUI-Custom-Nodes implements the stable-diffusion-x4-upscaler via DeepFloyd IF stage 3 nodes. I've just tried it, but not sure I've done it correctly as the results were soft and had halos around features.
That's not the same thing. How does that work with x4-upscaler-ema? You can't load the model as an upscaler, only a checkpoint. But that node requires both a checkpoint and an upscaler.
Zulini IF II and IF III nodes have some functionality for using the stable-diffusion 4x upscaler. I tried it, wasn't impressed with the results sadly.
The controlnet tile upscale only works with SD 1.5 since there isn't a controlnet XL tile model so far
https://github.com/Zuellni/ComfyUI-Custom-Nodes