Lates News

According to the Qwen large model, today it is open-sourcing Qwen-Image-2.1, an image model that balances generation quality, inference efficiency, and usage cost. Qwen-Image-2.1 integrates text-to-image generation and image editing into a single model. The visual generation component has only 7B parameters and natively supports the generation and editing of transparent images. The main features of this updated model include four aspects: 1. Small model, strong performance, ultimate cost-effectiveness: a lightweight model structure and inference optimization strike a balance between generation quality and computational cost. 2. Native transparency, integrated creation and editing: generate ordinary images or transparent images based on prompts, and support transparent layer editing and cutout. 3. All-around editing, multi-reference preservation: support up to 10 reference images, enhance local editing, portrait and product fidelity, and cover a variety of editing tasks. 4. Realistic texture, elegant text and beautiful people: improve text layout, character lighting and detail expression, making generated results more aesthetically pleasing. (Qwen large model).