Free · runs on a Hugging Face GPU, no sign-in needed
Qwen-Image-2.1-Turbo AI Image Generator
Qwen-Image-2.1-Turbo is the official 8-step distilled version of Alibaba's Qwen Image 2.1. It keeps the full model's legible text and detail, renders a 1K image in seconds, and can edit up to 3 reference images from a single instruction.
What is Qwen-Image-2.1-Turbo?
Qwen-Image-2.1-Turbo is an accelerated checkpoint of Qwen Image 2.1, released by Alibaba's Qwen team. It uses the same 7B visual generation architecture as the full model, but was distilled to produce a finished image in just 8 denoising steps.
Fewer steps means much faster results: a 1K image takes a few seconds of GPU time, and a 2K poster well under half a minute. Because the architecture is unchanged, Turbo keeps what makes Qwen Image 2.1 stand out — sharp, correctly spelled text, faithful people and products, and native transparency.
Turbo also handles image editing. Add up to three reference images and describe the change: restyle a photo, swap a background, or combine subjects from several pictures into one scene.
Why use Qwen-Image-2.1-Turbo
8 steps, seconds per image
The distilled schedule ships with the checkpoint and runs at CFG 1, so every image is fast without tuning samplers or step counts.
Text that reads right
Posters, packaging, infographics and study notes come out with clean, correctly spelled lettering, in English, Chinese and more.
Multi-image editing
Edit one photo or blend up to three. Refer to them as <image1>, <image2> and <image3> in your instruction.
Three resolutions
1K for quick drafts, 1.5K as a middle ground, and 2K — the model card's default — for the most detail.
Transparent backgrounds
Switch the background to Transparent for stickers, logos and cut-outs with a real alpha channel.
Prompt enhancement
Turn on enhancement and Qwen's official prompt rewriter expands a short idea into a detailed prompt before generating.
How to edit images with Qwen-Image-2.1-Turbo
Upload your images, then say what should change — the model keeps everything else.
- Number your references in the prompt: "Dress the woman in <image1> in the jacket from <image2>".
- Leave the aspect ratio on Auto to keep the shape of your first image, or pick a ratio to reframe it.
- One reference image runs at the resolution you choose; with two or three the resolution is lowered automatically so the edit fits in GPU memory.
What you can make with Qwen-Image-2.1-Turbo
Posters and social graphics
Layouts where the headline needs to be spelled right the first time.
Infographics and study notes
Dense, labelled diagrams and hand-lettered notes pages.
Photo edits
Change clothing, lighting, background or style while keeping the subject.
Composites
Bring people, products and places from separate photos into one image.
Stickers and cut-outs
Transparent PNGs ready for chat apps, merch and design files.
Fast iteration
Try ten ideas at 1K in the time a slower model makes one, then re-run the best at 2K with its seed.
How to use Qwen-Image-2.1-Turbo
- Write your prompt — or pick one of the ideas under the prompt box.
- Add reference images if you want to edit or combine photos; skip this to create from text.
- Choose the frame: an aspect ratio, a resolution and an opaque or transparent background.
- Click Generate and download the finished image from your history.
Frequently Asked Questions
Everything you need to know about generating images with Qwen-Image-2.1-Turbo.
- It is the official 8-step distilled checkpoint of Qwen Image 2.1 from Alibaba's Qwen team. It uses the same architecture as the full model and supports both text-to-image and image editing.
- Turbo generates in 8 steps, so it is much faster, it is free here, and it can edit reference images. The full model offers 21:9 and 9:21 frames, WebP and JPEG output, and is the one to use for commercial work: try it on our Qwen Image 2.1 page.
- Yes. Add up to three reference images (JPEG, PNG or WebP, 10 MB each) and describe the change, referring to each one by the label shown on it in the panel.
- 1K, 1.5K and 2K, in 1:1, 4:3, 3:4, 3:2, 2:3, 16:9 or 9:16. Auto follows your first reference image, or makes a square image without one.
- Yes — no credits and no sign-in. Images are generated on a public Hugging Face GPU straight from your browser, which gives every visitor a daily allowance of free GPU time. Editing, transparent backgrounds and prompt enhancement are free too.
- Turbo runs on shared Hugging Face GPUs. At busy times you may wait in a queue or time out — just try again, or pick 1K, which is fastest. If you have used up your free GPU time, it refills over time, so try again later. Edits whose prompt is flagged by the safety filter are also refused.
- Qwen-Image-2.1-Turbo is released under the Qwen Research License, which permits research and evaluation use only. For commercial projects, use Qwen Image 2.1 instead.