Qwen Image Edit

Advanced semantic and appearance-aware image editing defines Qwen-Image-Edit, combining Qwen2.5-VL’s visual understanding with a powerful VAE encoder for pixel-level control. It enables precise text changes, object edits, rotations, and stylistic transformations while preserving character consistency.

Try Qwen Image Edit

Examples of Qwen Image Edit

Four wizard character poses—singing, playing football, spellcasting, and reading a magic book—showcasing consistent character generation using Qwen Image Edit.

A red sports car edited into a matte-black version on the same street background, showing precise image editing with Qwen Image Edit.

Why You’ll Love It

Dual-Layer Control

Semantic + appearance editing for full creative flexibility.

Precise Text Edits

Modify bilingual text while preserving font and style.

Object Changes

Add, remove, or modify elements easily in seconds.

Consistent Characters

Maintain identity across edits, styles, and rotations.

Creative Transfer

Supports IP creation, style transfer, and full 180° rotations.

Benchmark Proven

State-of-the-art results in multiple editing benchmarks.

Trusted by Global Creators

Haruka T.

The text editing is unbelievably accurate. Fonts and styles stay perfect even after multiple changes.

Daniel F.

I love the semantic control. Rotations, style changes, and object edits always keep character identity intact.

Giulia R.

It’s the first model that handles appearance and meaning together. Great for creating branded visuals and IP variations.

How Qwen Image Edit Works

Step 1

Upload Image

Choose an image to edit with Qwen-Image-Edit.

Step 2

Add Instructions

Specify text edits, semantic changes, or appearance adjustments.

Step 3

Generate

Qwen applies dual-model processing for precise, consistent edits.

Try Qwen Image Edit

Frequently Asked Questions

An advanced image editing model built on Qwen-Image with dual semantic and appearance control.

Yes — it supports direct bilingual (Chinese + English) text editing while preserving font, size, and style.

Yes. You can add, remove, or modify elements while keeping other regions untouched.

It supports semantic edits including 90° and 180° rotations for new-view synthesis.

Yes — character identity and core semantics remain stable across edits and style changes.

The model reaches state-of-the-art performance in public image editing benchmarks.