Alibaba releases Qwen-Image-2.1 with transparent background support
Alibaba's Qwen team released Qwen-Image-2.1 on September 20, 2026. This image generation model combines creation and editing in a single system, allowing users to generate images with transparent backgrounds directly. This feature removes the need for separate post-processing steps to cut out subjects for stickers, product shots, or video overlays.
Key points
- Qwen-Image-2.1 generates images with transparent backgrounds directly, eliminating post-processing cutouts.
- The model supports up to 10 reference images for combining elements like outfits or room layouts.
- Weights are downloadable but licensed for research only; commercial use requires a separate contract.
The model supports up to 10 reference images simultaneously, enabling users to combine elements like clothing, accessories, or furniture into a single coherent scene. It also allows for partial editing, such as changing a background while keeping a specific product or person intact. The core architecture uses 7 billion parameters, with optimizations that enable the generation of 2K-sized images.
While the model weights are available for download on Hugging Face and ModelScope, the license is restricted to research and evaluation. Commercial use requires a separate contract with Qwen. This distinction highlights that open availability does not equal commercial freedom, a factor increasingly influencing how businesses select AI tools for production workflows.
Model page: Qwen Image 2.1 →
[AI News Commentary] Qwen-Image-2.1 Released: Creating Images with Transparent Backgrounds Directly, AI Images Moving Closer to 'Assets'
note.com · 21 September 2026
Loading the full article…
This text was published by note.com and written by うさとものひとり言. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Generative AI & Models
All →- GPT-6 Astra scores 99.9% on ARC-AGI-3, OpenAI says · 5 src
- Anthropic releases Claude Opus 5.5 with 40% lower costs · 9 src
- Meta's Muse agent gives every user a full cloud computer running Ubuntu Linux · 6 src
- Liquid AI releases LFM2.5-VL-DSpark for faster vision-language model inference · 2 src
- Google releases Gemini 3.8 Flash TTS and Flash-Lite TTS with voice cloning and 100+ languages · 11 src
Comments
via GitHub Discussions