Qwen-Image handles long, detailed natural-language prompts and supports a separate negative prompt, which makes it easy to steer away from unwanted artifacts.
The generator writes a subject-first description followed by scene and lighting, and outputs your negatives as their own field.
Describing the subject first, then the setting, then lighting and camera keeps the result coherent.
Add realism-related negatives such as “plastic skin” or “CGI” for photographic results; the generator suggests them automatically.
For changing an existing photo, use the edit model and describe only the change you want.
Qwen-Image is known for strong text rendering. Add the exact text in the notes field and describe where it should appear.
Plain text is the default; you can switch to JSON in the builder if your workflow prefers it.