Instructions to use joyfox/Qwen-Image-Edit-2511-Upscale2K with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use joyfox/Qwen-Image-Edit-2511-Upscale2K with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline from diffusers.utils import load_image # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("Qwen/Qwen-Image-Edit-2511", dtype=torch.bfloat16, device_map="cuda") pipe.load_lora_weights("joyfox/Qwen-Image-Edit-2511-Upscale2K") prompt = "Upscale this picture to 4K resolution. " input_image = load_image("https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/cat.png") image = pipe(image=input_image, prompt=prompt).images[0] - Inference
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- Draw Things
Which Settings Are Required for Strict, Identity-Preserving Upscaling with Qwen Image Edit?
I’ve installed everything, but when I try to upscale an image, it produces results that are different from the original image instead of actually upscaling it. If anyone has managed to get proper upscaling working, could you share which settings you are using?
I’ve installed everything, but when I try to upscale an image, it produces results that are different from the original image instead of actually upscaling it. If anyone has managed to get proper upscaling working, could you share which settings you are using?
It seems hard to balance the generation-ality and reconstructioin for QwenImageEdit series model. I tried 2509 and some similar phenomenon found.
The solution can be referred to in this video:https://www.bilibili.com/video/BV14gBSBVEPT/?share_source=copy_web&vd_source=237923f28060115b2ed44346e2c1e48d
The solution can be referred to in this video:https://www.bilibili.com/video/BV14gBSBVEPT/?share_source=copy_web&vd_source=237923f28060115b2ed44346e2c1e48d
the code show in model card page gives no good results as shown on that page. Maybe comfyUI needed?