Qwen-Image Edit

Yes, Qwen-Image Edit is built directly on the Qwen-Image-Edit-Plus and Qwen2.5-VL 20-billion parameter vision-language architecture developed by Alibaba Group. It operationalizes this foundation with specialized LoRA modules optimized for specific image tasks—including background staging, garment graphics mapping, headshot expansion, and cinematic frame continuity.

Choose Add Background when you have an isolated subject on a plain or white background and want the model to generate a brand-new environment from a text prompt. Choose Integrate Product when you already have two specific images—a product and a distinct background photograph—and need the model to blend the product into that target scene with matching 3D perspective, reflections, and contact shadows.

Use the Shirt Design operation when mapping a 2D graphic onto clothing, providing the flat artwork as the first image and the model photo as the second image while using the prompt trigger put this design on their shirt. For transforming a tight face crop into an upper-body or full-body editorial image with custom styling, select the Face to Full Portrait operation.

Select the Next Scene operation and start your prompt with the mandatory prefix Next Scene: followed by your intended camera motion or environmental evolution. For the smoothest cinematic transition, set the effect_strength slider between 0.7 and 0.8 to balance narrative progression with visual scene continuity.

The model accepts up to 4 images depending on the selected task. Shirt Design requires exactly 2 images (the flat graphic and the apparel photo), Group Photo accepts between 2 and 4 individual headshots to blend into a unified composition, and single-subject operations like Add Background or Remove Element take 1 image.

The model is not designed for pixel-perfect virtual try-on where intricate garment stitching or exact proprietary patterns must match 1:1, nor does it generate images purely from text without input imagery. It also struggles with micro-adjustments to individual fingers and ultra-fine latent inpainting, where manual brush masks remain more effective.