Introducing the image generation and editing AI 'Qwen-Image-Edit-2511,' which improves consistency of people and objects and incorporates the popular LoRA algorithm for even higher quality.



The Qwen series of AI models from Alibaba, a leading Chinese technology company, has released ' Qwen-Image-Edit-2511, ' an enhanced version of

Qwen-Image-Edi, an image generation AI equipped with image editing tasks. It can maintain consistency of people even in group photos and complex scenes, and its built-in LoRA technology expands its expressive power without the need for additional adjustments.

Qwen-Image-Edit-2511: Improve Consistency
https://qwen.ai/blog?id=qwen-image-edit-2511






Qwen-Image-Edit-2511 is a significantly enhanced version of Qwen-Image-Edit-2509 , released in September 2025. While retaining the fundamental features of Qwen-Image-Edit, such as language-based image editing and high-quality consistency, it offers significantly improved performance in preserving a person's face and features during the editing process. Below is an example of editing provided by Alibaba, where the input image on the left is edited into the image on the right, featuring different clothing and poses, using only natural language editing prompts. Even when editing the same person in different poses and backgrounds, editing without compromising the original features is possible.



It also makes it easier to reflect people's characteristics when creating character illustrations.



While Qwen-Image-Edit-2509 already had high consistency for single-subject photos, Qwen-Image-Edit-2511 significantly improves the consistency of group photos of multiple people. It also excels at editing two separate photos of people, or combining a photo of a person and a cat.



In addition to inheriting the high-precision image editing capabilities from its predecessor, Qwen-Image-Edit-2511 also boasts the integration of the popular Low-Rank Adaptation (LoRA) into its model. LoRA is a lightweight extension technology that allows users to retroactively learn and add specific artistic styles, expressive habits, and subject depiction techniques without rewriting the underlying large-scale language model. Many LoRa extensions for Qwen-Image-Edit created by volunteers are available online. Because Qwen-Image-Edit-2511 integrates multiple popular LoRa extensions, users can use LoRa-based features such as lighting control without having to install LoRa themselves.



Qwen-Image-Edit-2511 focuses on editing accuracy and structural expression for product design and industrial applications, improving the consistency of industrial design images by generating multiple directions of a single product while maintaining consistency, and exchanging the color and material of complex products.



Additionally, Qwen-Image-Edit-2511 introduces more powerful geometric reasoning capabilities, allowing users to directly generate auxiliary construction lines for design and annotation purposes, such as editing architectural and design elements.



A demo of Qwen-Image-Edit-2511 is available at Hugging Face.

Qwen Image Edit 2511 - a Hugging Face Space by Qwen
https://huggingface.co/spaces/Qwen/Qwen-Image-Edit-2511



Related Posts:

in AI, Posted by log1e_dh