Qwen-Image-Edit-2509

Overview

Qwen-Image-Edit-2509 is an advanced 20-billion parameter multimodal image editing foundation model by QwenLM. It excels at editing and blending multiple input images with outstanding context preservation, face identity retention, and object coherence. Its unique capability includes pixel-level ControlNet-based control for pose and scene manipulation.

Technical Specifications

Qwen-Image-Edit-2509 is a 20-billion parameter multimodal diffusion model architecture, optimized for image editing and multi-image fusion. It supports input of 1 to 3 RGB images along with ControlNet-based conditioning maps such as depth, edges, and keypoints for fine pose and structure control. The model enhances single-image consistency drastically, retaining facial identity, product details, and original text styles including fonts and colors.

Performance Benchmarks

• Outperforms leading image editing models in identity consistency tests for faces and products.
• Excels in multi-image blending tasks by maintaining semantic and spatial coherence.
• Demonstrates superior quality in text editing robustness, preserving text style across edits.
• Consistently scored higher in user evaluations compared to Nano Banana and other commercial competitors in multi-image and complex text-image editing benchmarks.

Key Features

• Multi-Image Editing Support: Seamlessly edit and merge up to 3 input images, including combinations like person+person, person+product, and person+scene.
• Enhanced Single-Image Consistency: Significant improvements in maintaining facial identities through pose changes, portrait style variations, and preserving product identities.
• Advanced Text Editing: Modify text content in images with control over font type, color, and material textures—supporting precise text rendering and integration with visual edits.
• Bilingual Capabilities: Supports both Chinese and English text editing natively.

Use Cases

• Professional photo editing with complex multi-person or scene compositions
• Advertising creatives requiring product poster consistency
• Digital art creation with precise pose and layout control via ControlNet
• Meme and text-based image generation preserving style continuity
• Restoring old photos while maintaining identity and context
• Comprehensive content creation for marketing and social media

Comparison with Other Models

vs Gemini 2.5 Flash Image: Qwen excels in precise text editing and multi-image editing with native ControlNet support; Gemini leads in photorealistic 3D-style character rendering and multi-step prompt understanding. Qwen supports bilingual text editing (Chinese and English); Nano Banana’s text editing capabilities are limited.

vs Stable Diffusion 1.5: Qwen-Image-Edit-2509 supports multi-image inputs and ControlNet keypoint control, Stable Diffusion mainly single-image edits with some ControlNet capabilities. Stable Diffusion is more lightweight and has larger community ecosystem; Qwen has superior editing precision but requires larger GPU resources.