Views
No views yet
[!note] This adapter was experimental and trained to recover facial patterns from partially blurred images. The base model produces better overall quality, but it does not preserve those patterns.
[!note] Note:prithivMLmods/Qwen-Image-Edit-2511-Unblur-Upscalemay make mistakes. Please check the Important Info section.
Qwen-Image-Edit-2511-Unblur-Upscale is an adapter LoRA developed for Qwen’s Qwen-Image-Edit-2511 image-to-image model, specifically designed to unblur and upscale images to high resolution. The model enhances image clarity by reducing blur, restoring fine details, and improving overall sharpness while preserving natural textures and realistic colors. It maintains strong visual consistency across lighting, edges, and color tones, ensuring the upscaled output looks clean, natural, and photographically authentic rather than over-processed. This makes it well suited for restoring low-quality, blurry, or low-resolution images into sharp, high-fidelity results. Above are example outputs demonstrating the model’s ability to deliver clear, high-resolution, and visually cohesive enhancements.
Compatible with versions 2509 and 2511.
pip install -U diffusers transformers accelerate1import torch
2from diffusers import DiffusionPipeline
3from diffusers.utils import load_image
4
5# Switch to "mps" for Apple devices
6pipe = DiffusionPipeline.from_pretrained(
7 "Qwen/Qwen-Image-Edit-2511",
8 dtype=torch.bfloat16,
9 device_map="cuda"
10)
11
12pipe.load_lora_weights("prithivMLmods/Qwen-Image-Edit-2511-Unblur-Upscale")
13
14prompt = "unblur and upscale"
15
16input_image = load_image(
17 "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/cat.png"
18)
19
20image = pipe(image=input_image, prompt=prompt).images[0]Recommended: Qwen-Image-Edit-Unblur-Upscale_20.safetensors / Qwen-Image-Edit-Unblur-Upscale_15.safetensors
unblur and upscale