Qwen-image-2.1_loras

Qwen-image-2.1_loras

Adds two image-editing LoRAs to Qwen-Image-2.1, one that fixes pixel shift and one that removes a masked area. Suits people who edit images with that base model.

Made by GooHAI on Hugging Face. These notes are Mindket's, written from its model page and model card; the model's own files and license are on Hugging Face.

About

What it is: A small collection of two image-editing LoRAs from GooHAI, trained on top of Qwen-Image-2.1. One fixes the pixel shift that edits with Qwen 2.1 can show. The other removes whatever sits inside a masked area, with no shift and no color change.

Who it suits: People who edit images with Qwen-Image-2.1 and want cleaner results, such as retouchers, product photographers and designers.

Industry | Who uses it | How they use it | Example
Photography | Photo retouchers | Remove an unwanted object inside a mask | Clean a distracting item out of a picture
E-commerce | Product photo editors | Edit a picture without the content shifting | Keep a product aligned after an edit
Design | Designers | Apply a LoRA at a set scale in an edit workflow | Run the shift fix at 0.7 on every edit

If you do not use the Qwen-Image-2.1 base model, you can skip it.

Where to use it:

• Any image-editing setup that loads Qwen-Image-2.1 and safetensors LoRAs
• Windows, macOS or Linux, for downloading the files

Use case | Who | What they get back
Fix pixel shift | Editors | Edits that keep pixels in place
Remove masked area | Retouchers | The masked region removed, without color deviation

How to set it up:

• Python with pip, to download the files
• The base model Qwen-Image-2.1
• A setup that can load LoRA files

1. Install the Hugging Face hub package.
2. Download the repo.
3. Load the base model, then add the LoRA file you want.

pip install -U huggingface_hub
hf download GooHAI/Qwen-image-2.1_loras

How to use it: Load the base model, load the matching LoRA at its recommended scale, and add the trigger phrase if there is one. The trigger phrase for the removal LoRA is Chinese text.

移除图像中的绿色区域

Tool | What it does
Consistency no-offset GH_V1.0 | Fixes the pixel offset problem of Qwen 2.1, scale 0.7, no trigger word
Remove anything Remove_GH | Removes the masked region without offset or color deviation, scale 1.0, uses the trigger phrase above

Example: You mark an unwanted object with a mask, load the removal LoRA at scale 1.0 and use the trigger phrase, which asks the model to remove the green region of the image. The masked area is removed and the rest of the picture stays in place.

For newcomers and experts: If you are new, a LoRA is a small add-on file that changes how a base model behaves. Load Qwen-Image-2.1 first, then add one LoRA at the scale the card recommends.

Both files are image-editing LoRAs fine-tuned on Qwen-Image-2.1. The card gives scales of 0.7 for the shift fix and 1.0 for removal, and shares no training details.

Load the Qwen-Image-2.1 base model
Load a LoRA at its recommended scale
Add the trigger phrase if one applies
Generate or edit the image

Good to know:

• The card is bilingual, in Chinese and English
• The shift fix needs no trigger word
• License: Apache 2.0
• Cost to run: none named
• Data it reaches: the images you edit and the files downloaded from Hugging Face
• Not for: use without the Qwen-Image-2.1 base model

On Hugging Face, 13 November 2026

• Task: image to image
• Downloads in the last 30 days: 0
• Likes: 17
• License: apache-2.0
• Based on: Qwen/Qwen-Image-2.1
• Tags: qwen-image, lora, image-editing, image-removal, image-to-image

Links

• Model: https://huggingface.co/GooHAI/Qwen-image-2.1_loras
• Files: https://huggingface.co/GooHAI/Qwen-image-2.1_loras/tree/main

Get the model

Download its files with Hugging Face's command line:

pip install -U huggingface_hub
hf download GooHAI/Qwen-image-2.1_loras

Then follow the usage in its model card. Nothing in this kit runs by itself.