RU EN
Log in Open the studio

Images · Text → image · Image → image

Qwen Image 3

Rich layouts, infographics, and small in-image text in 12 languages — in Standard and Pro

Open the studio

About the model

Qwen Image 3 is the next generation of Alibaba’s model. Compared with the previous version, it follows complex instructions more precisely, renders realistic detail better, and draws on broader world knowledge. On NanoBananana it comes in two modes — text-to-image and image-to-image — and two versions: Standard for everyday work and Pro for more demanding delivery.

The model’s main strength is rich, structured imagery. It can fit a newspaper page, a storyboard, an exam sheet, or a detailed infographic into a single frame, arranging elements logically and keeping even small text legible. Text is rendered natively in 12 languages, which makes it a good fit for multilingual materials.

In image-to-image mode you can upload up to three sources and describe the edit in words: change the style, add an object, or move a character into a new scene. Built-in prompt enhancement helps you get a solid result even from a short description.

Strengths

  • Complex structured compositions — newspaper pages, storyboards, infographics, learning materials
  • Legible small text and clean typography right inside the image
  • Native text rendering in 12 languages
  • Realistic detail — skin texture, hair strands, material surfaces
  • Prompt-driven editing based on one to three source images
  • 1K or 2K output and eight aspect ratios, including 21:9

Best for

  • Infographics, reports, and newspaper- or magazine-style pages
  • Storyboards for videos, ads, and animation
  • Mockups of websites, mobile apps, and game interfaces
  • Posters, banners, and product cards with captions
  • Learning materials — diagrams, slides, worksheets

Limitations

  • Maximum resolution is 2K
  • Up to three source images, 10 MB each
  • Prompts are limited to 5,000 characters

Prompting tips

  • Describe the layout block by block — headline, columns, illustrations, captions
  • Put in-image text in quotes and state the language
  • The model handles long, detailed prompts well — feel free to list every element
  • In image-to-image mode, say what to keep first, then what to change
  • For final delivery and small text, pick the Pro version at 2K

Other models in the service that fit similar tasks — try them in the same studio.