Native RGBA output
Generate color and alpha together instead of flattening every result onto a background.
Official model releasedCommercial API pending
Qwen Image 2.1 Transparent PNG Generator
Prepare native RGBA images from a prompt or product upload, inspect transparency on different backgrounds, and explore the workflow while commercial API rights are being verified.
This is a prepared product workflow, not a live Qwen Image 2.1 generation endpoint. It does not upload to Qwen, create a paid task, or charge credits.
Site-owned preview · not a Qwen output
Product cutout
Transparent PNG workspace
The interaction is ready; commercial model execution stays locked.
Choose a transparent asset type
Prepared output settings
- Output
- Transparent PNG
- Color
- RGBA
- Resolution
- Native 2K
- References
- Up to 10 documented
No provider request is sent and no credits are charged from this preview.
Use the live background removerTransparent result preview
Switch the matte to inspect real alpha edges in the site-owned examples.
Officially documented capabilities
Why native transparency is different
The official release describes one pipeline for regular and transparent generation, transparent-layer editing and subject extraction. Hosted API support may expose a smaller contract, so provider limits will be verified separately.
Up to 10 references
The official model supports multi-reference composition with as many as ten visual inputs.
Local mask editing
Circles, painted annotations and separate masks can specify the regions that should change.
Subject consistency
The release highlights identity preservation for people and products across edits.
Choose the right workflow
Native RGBA versus background removal
| Capability | Native RGBA generation | Background remover |
|---|---|---|
| Starting point | A prompt, reference image or transparent layer. | An existing photo with a visible background. |
| Alpha creation | Color and transparency can be created together by the model. | A segmentation model estimates the subject after the photo already exists. |
| Best fit | New stickers, design layers, assets and transparent products. | Fast cutouts from product, portrait and catalog photos. |
| Availability here | Prepared preview; commercial execution remains locked. | Available now through the dedicated background-removal tool. |
Reusable transparent assets
One alpha channel, many workflows
Transparent output is most useful when the subject must move between layouts without repeating the extraction step.
Ecommerce product cutouts for catalogs and marketplaces
Sticker and mascot assets for messaging and social posts
Transparent logo concepts and brand symbols
Game characters, props and interface assets
Reusable overlays for Photoshop, Canva and video editors
Transparent lighting, glass and soft-edge design layers
Commercial use is not enabled yet
Qwen Image 2.1 weights are published under the Qwen Research License. The agreement limits the supplied materials to non-commercial use unless a separate commercial license is obtained. A hosted provider must also publish or confirm terms that permit this SaaS use before generation can be enabled here.
FAQ
Qwen Image 2.1 transparency questions
Can Qwen Image 2.1 generate a real transparent PNG?
The official release documents native RGBA generation and saves transparent outputs as images with an alpha channel when transparency is produced.
Is native RGBA the same as removing a background?
No. Native RGBA can create color and transparency together during generation. Background removal starts with an existing image and estimates which pixels belong to the subject.
Can I use Qwen Image 2.1 commercially?
The published model weights use the Qwen Research License, which limits the materials to non-commercial use unless a separate commercial license is obtained. Hosted API terms must be checked independently.
How many reference images does the model support?
The official model documentation states support for up to ten reference images. A future hosted API may expose a different limit, which will be verified before launch.
Can it extract a product from a normal photograph?
Subject extraction from photographs is listed as part of the unified transparency workflow. The live Qwen endpoint is still disabled here, while the separate background remover is available now.
Does this preview charge credits or upload my file?
No. The preview runs only local interface interactions. It does not create a provider task, upload to Qwen or consume site credits.
Explore Qwen Image 2.1 without mixing search intent
Use the model page for the broader generation, editing and multi-reference workflow, or use the live remover when you need a transparent cutout today.