QwenLM/Qwen-Image-2.1 ? reverse-engineered prompt

Build me a simple local app around Qwen Image 2.1 that lets me create images from a text prompt and also edit an existing image.

I want a clean interface where I can type what I want, upload one source image if I am editing, and optionally add a mask or marked area for local changes. If the model supports transparent images, include a toggle for that too, so I can generate regular images or transparent ones. It would be great if I can also add a few reference images to help keep a person or product looking the same.

Please make it easy to run from one command, with a nice example page and clear defaults that work out of the box. If you need to check the latest docs for the model or the image library, go ahead and look them up online. Keep the result polished, fast to use, and focused on image generation and editing rather than extra fluff.