Qwen Image 2.1 Studio

Qwen Image 2.1 Studio

Bundles nine image tools for Qwen Image 2.1 in one Gradio app, from generation and editing to relighting and outfit swapping. For people who want to try or host these tools.

Made by Maximofn on Hugging Face. These notes are Mindket's, written from its Space page and README; the app's own files and license are on Hugging Face.

About

What it is: Qwen Image 2.1 Studio is a Gradio app that bundles nine image tools for the Qwen Image 2.1 model in a single page. Its tools cover making and editing pictures, removing or moving objects, changing the lighting and viewpoint, and extending a picture past its edges. The interface opens in Spanish, with a switch to English.

Who it suits: It suits designers, content makers and developers who want several Qwen Image 2.1 recipes in one place, and people who host their own image tools.

Industry | Who uses it | How they use it | Example
Design | Designers | Edit and extend pictures with several Qwen Image 2.1 tools | Widen a picture with the outpainting tab
Content creation | Creators | Fix or restyle photos without switching apps | Remove a distracting object from a photo
Fashion and retail | Small shop owners | Try outfit changes on a picture | Swap the outfit in a model photo
Software | Developers | Host the studio on their own GPU Space | Copy the Space and set its hardware

If you want a plain one-prompt image generator, you can skip it, since this app is built around many separate tabs.

Where to use it:

• A web browser, on the Space's page on Hugging Face
• A local copy of the files, run on Python 3.12 on a machine with a suitable GPU
• Your own Hugging Face Space set to ZeroGPU or a GPU with enough VRAM
• A self-hosted setup, where the page keeps a local model list

Use case | Who | What they get back
Edit a photo | Creators | A changed picture, such as one with an object removed or the light adjusted
Extend a picture | Designers | A larger canvas made by outpainting
Host the studio | Developers | The nine tools running on their own hardware

How to set it up:

• A web browser, which is all the online page needs
• Python 3.12 and the packages in requirements.txt, for a local copy
• ZeroGPU, or a GPU whose memory can hold the BF16 base model, the official Turbo generator and the adapters
• A Hugging Face sign-in if you want the workflow nodes to use your own quota

1. Open the Space's page on Hugging Face to use it online. Nothing needs installing.
2. For a local copy, install the Hugging Face hub package and download the Space's files.
3. Move into the downloaded folder.
4. Install Gradio 6.28.0, then the packages from requirements.txt.
5. Start the app with python app.py. The first start downloads the pinned base model, the Turbo parts and the LoRA files.

pip install -U huggingface_hub
hf download Maximofn/Qwen-Image-2.1-studio --repo-type space --local-dir Qwen-Image-2.1-studio
cd Qwen-Image-2.1-studio
pip install gradio==6.28.0
pip install -r requirements.txt
python app.py

How to use it: Open the page and pick the tab for the job. Each tab keeps its own model choice, and a request sets its adapters explicitly, so one tool's LoRA does not leak into another. The cloud model list has 17 profiles in all.

Tool | What it does
Generation and editing | Makes new pictures and edits existing ones
Workflow nodes | Runs node-based image workflows, with Hugging Face sign-in for your own quota
Object removal | Takes an object out of a picture
Viewpoint orbit | Shifts the viewpoint on a picture
Moving objects | Moves objects around inside a picture
Face swapping | Swaps faces in pictures
Relighting | Changes the lighting of a picture
Outpainting | Extends a picture beyond its edges
Outfit swapping | Changes the clothes in a picture

Example: Open the object removal tab, upload a photo and choose the object to take out. The app runs the request on the GPU and returns the photo without that object.

For newcomers and experts: Open the Space page and try one tab with a harmless picture. The language switch is there if you would rather read English, and private photos are better kept out of a public demo.

On ZeroGPU the base BF16 pipeline, the named LoRAs and the separate official Turbo checkpoint start once, and the frozen text encoder is shared after all 750 tensors were compared byte for byte. Each request picks the right pipeline and turns on only its adapters. A Space on CPU hardware opens the editor with a note that no GPU is available, and the server blocks generation. FP8, NVFP4 and GGUF profiles belong to local engines and are left out.

You pick a tab and add a picture
The app picks the pipeline for that tab
Only that tool's adapters are switched on
A GPU is allocated for the request
The result appears in the page

Good to know: Generation on ZeroGPU spends the visitor's ZeroGPU quota, and results live in temporary storage that a restart clears.

• License: the studio code is Apache-2.0, while model weights keep their own licenses and the Qwen research license restricts commercial use
• Cost to run: no outside inference service or private-server login is needed, only ZeroGPU or a GPU with enough memory, and large downloads on the first start
• Data it reaches: the pictures you upload, which are processed inside the Space, and model files downloaded from Hugging Face
• Not for: commercial use of outputs before reading the linked model cards, or private photos on a public demo

On Hugging Face, 13 November 2026

• Likes: 8
• Built with: gradio
• License: apache-2.0

Links

• Space: https://huggingface.co/spaces/Maximofn/Qwen-Image-2.1-studio
• Files: https://huggingface.co/spaces/Maximofn/Qwen-Image-2.1-studio/tree/main

Try it

Open the Space in your browser, or copy it to your own Hugging Face account from its page. Nothing in this kit runs by itself.