Available now

Qwen Image 3.0: image editing

By Alibaba

Alibaba's Qwen Image 3.0 turns a long prompt into one image at 1K or 2K, and edits your photos from a written instruction using one to three images.

Model ID
Maker
Alibaba

Input

Fields marked * are required. Review the price before generating.

Text prompt describing the desired edit. Supports Chinese and English. Max 5000 characters.

Example prompt. Edit it or write your own.

This model reads prompts in English or Chinese.

If not provided, the model determines the resolution automatically. Total number of pixels must be between 512x512 and 2048x2048.

Reference images for editing (1-3 images required). Order matters: reference as 'image 1', 'image 2', 'image 3' in prompt. Resolution: 384-2048px each dimension. Max size: 10MB each. Formats: JPEG, JPG, PNG (no alpha), WEBP.

Uploading a file needs an account. Create account

JPEG, PNG, WebP · up to 25 MB

Additional settings

Content to avoid in the generated image. Max 500 characters.

Enable automatic LLM prompt rewriting for better results.

Random seed for reproducibility (0-2147483647).

Price$0.12 / run

You sign up first. Your settings are kept.

About this model

Overview

The text mode takes a prompt of up to 4500 characters and renders one image in eight shapes from 21:9 to 9:16. Prompt extend rewrites a short prompt into a fuller one, and thinking mode, which only works with prompt extend on, takes longer to plan the image.

The edit mode takes one to three images, called image 1, image 2 and image 3 in the order you added them, and an instruction in English or Chinese. Each run returns up to 6 images as PNG, JPEG or WebP; a negative prompt lists what to keep out.

What you can build

  • Detailed scenes from a long written brief
  • Wide 21:9 headers and tall 9:16 phone images
  • Outfit or product swaps between two photos
  • Small fixes: remove an object, change a color, add a sign

Prompting tips

  • Turn on prompt extend for one-line ideas; keep it off when the wording must stay as you wrote it.
  • Thinking mode is on by default but has no effect until prompt extend is on.
  • In the edit mode, name the images by number in the order you uploaded them; the order changes the result.
  • Say what must stay the same, such as the face, the pose or the product label.

What to check

Read every letter and check alignment, spacing and the placement of objects.

Inputs

  • Prompt (required): Text prompt describing the desired edit. Supports Chinese and English. Max 5000 characters.
  • Negative Prompt (optional): Content to avoid in the generated image. Max 500 characters.
  • Image Size (optional): The size of the generated image. If not provided, the model determines the resolution automatically. Total number of pixels must be between 512x512 and 2048x2048. Options: Square HD, Square, Portrait 4:3, Portrait 16:9, Landscape 4:3, Landscape 16:9.
  • Enable Prompt Expansion (optional): Enable automatic LLM prompt rewriting for better results.
  • Seed (optional): Random seed for reproducibility (0-2147483647).
  • Number of Images (optional): The number of images to generate. From 1 to 6.
  • Output Format (optional): The format of the generated image. Options: jpeg, png, webp.
  • Images (required): Reference images for editing (1-3 images required). Order matters: reference as 'image 1', 'image 2', 'image 3' in prompt. Resolution: 384-2048px each dimension. Max size: 10MB each. Formats: JPEG, JPG, PNG (no alpha), WEBP.

Output and limits

  • Image sizes: Square HD, Square, Portrait 4:3, Portrait 16:9, Landscape 4:3, Landscape 16:9.
  • File formats: JPEG, PNG, WEBP.
  • Up to 6 images per run.
  • Small or dense text in the image can come out misspelled.
  • Hands, crowds and complex poses can show artifacts; generate a few variations and pick the best.
  • Some prompts or outputs can be refused by safety checks.

Price guide

  • up to, default options: $0.12 / run