Learn more about the Qwen-Image series. Try on Qwen Chat, or download model to run locally with ComfyUI or diffusers.
This version does not include the 'prompt polisher', and defaults to 16 inference steps to target ~30s generation times.