Alibaba QwenLM • ZeroGPU Ready

Qwen-Image-2.1

Unified Text-to-Image Generation & 10-Image Reference Conditioning with Native 2K Output, Transparent (RGBA) Support, and Integrated Prompt Enhancement.

Automatically expands prompts before generation.

Default off: Fast mode (8s). On: Full 8-step CoT (60s).

Upload 1 to 10 reference images. In your prompt, refer to them as Image 1, Image 2, etc. to guide composition, subject consistency, or style transfer.
Aspect Ratio
Resolution Preset

Native alpha transparency for stickers, icons, and graphic design assets.
15 60

Ready to generate.

📌 Key Capabilities:

  1. Prompt Enhancement (PE-T2I & PE-I2I):
    • Click ✨ Enhance Prompt or enable Auto-Enhance on Generate to automatically expand short prompts into rich, vivid descriptions with optimal scene details, lighting, and textures.
    • Fast Mode (Default, ~8s): Instantly rewrites the prompt directly.
    • Deep Thinking (~60s): Enables full 8-step Chain-of-Thought reasoning before outputting the rewritten prompt.
  2. Text-to-Image (T2I):
    • Generates ultra-high quality 2K images with realistic lighting, textures, and accurate text rendering.
  3. Multi-Image Reference (Up to 10 Images):
    • Refer to images as [Image 1], [Image 2], ... [Image 10] or first image, second image.
    • Subject consistency: A portrait of the person in [Image 1] wearing medieval knight armor in a forest.
    • Multi-subject fusion: A photo of the person in [Image 1] and the person in [Image 2] sitting together in a cafe.
    • Style Transfer: A futuristic city in the artistic watercolor style of [Image 3].
  4. Transparent (RGBA):
    • Enable the RGBA checkbox to create stickers, game sprites, or design assets with transparent backgrounds.

🌟 Curated Examples
Examples
Prompt Negative Prompt Aspect Ratio Resolution Preset Inference Steps Seed ✨ Generate Transparent Image (RGBA) Auto-Enhance on Generate 🧠 Deep Thinking (~60s)