Released Qwen-Image-Edit-2509_clear! Superior Texture Quality Over nano-banana
- Qwen-Image-Edit-2509_clear is now available.
- Superior texture and vivid contrast.
- Completely free commercial use.
Introduction
Hello everyone, I'm Easygoing.
Today, I'm excited to announce the release of the Qwen-Image-Edit-2509_clear model.
What is Qwen-Image-Edit-2509?
Qwen-Image is an AI image generation model released by China's Alibaba Group.
gantt
title Image Editing AIs
dateFormat YYYY-MM-DD
axisFormat %Y-%m
section Open AI
GPT-Image-1 : done, 2025-04-23, 2025-11-21
section Black Forests Labs
Flux.1-Kontext: done, 2025-06-26, 2025-11-21
section Google
nano-banana : done, 2025-08-26, 2025-11-21
section Alibaba
Qwen-Image-Edit : done, 2025-08-17, 2025-11-21
Qwen-Image-Edit-2509 : 2025-09-22, 2025-11-21
Qwen-Image-Edit is a model that supports image editing based on text instructions, and Qwen-Image-Edit-2509 is the latest version, released in September 2025.
Qwen-Image-Edit-2509 is a rival to Google's nano-banana image editing model, which I introduced previously, but it offers two significant advantages:
- It runs locally and can be used under a completely free license.
- No worry about the output being deleted by Google.
Enhancing the Texture of Qwen-Image-Edit-2509!
Due to the difference between online and local deployment, Qwen-Image-Edit-2509 and nano-banana vary significantly in model size and training data volume.
Because of its massive scale, nano-banana outperforms Qwen-Image-Edit in many areas, such as prompt adherence and the variety of illustrations. However, the expression of texture is arguably the most critical aspect in practical use.
Therefore, I've made adjustments to Qwen-Image-Edit-2509 to maximize its texture quality.
Comparing Illustrations!
Let's look at a comparison of the actual illustrations. On the left is the output from my adjusted model, Qwen-Image-Edit-2509_clear, and on the right is the output from the original Qwen-Image-Edit-2509 model.
Castle in a Dream Kingdom
White Blazer
Close-up View
The Qwen-Image-Edit-2509_clear model has been tuned for higher contrast and more vivid colors.
Even in the close-up of the eyes, the Qwen-Image-Edit-2509_clear model on the left shows a clearer representation and more accurate details.
Comparison with nano-banana in Image Editing!
Next, let's compare it with nano-banana using an image editing task.
Original Image
Colorization & Photo Restoration
The Qwen-Image model exhibits better photo texture.
Japanese Anime Character Style
Full-Body Shot
Character Sheet (Three-View)
Qwen-Image-Edit-2509_clear surpasses nano-banana in texture for photorealistic illustrations.
In anime illustrations, nano-banana better captures the essence of Audrey Hepburn, showing a deeper familiarity with her, but Qwen-Image-Edit-2509_clear's quality is by no means inferior.
For full-body shots and character sheet creation, both models achieve a practical level of quality.
Furthermore, since Qwen-Image-Edit-2509_clear is a model that can be used locally, the ability to try multiple generations by changing the seed value, coupled with its completely free license, offers enormous practical benefits.
Notes on Qwen-Image-Edit-2509_clear
While Qwen-Image-Edit-2509_clear has improved texture compared to the original Qwen-Image-Edit-2509, there are a few points to note.
Limited Compositional Variation
Qwen-Image is inherently a model that struggles with a wide variety of compositions.
Compared to the original Qwen-Image-Edit-2509, the Qwen-Image-Edit-2509_clear model has increased stability, but this comes at the cost of even less compositional variation.
This shouldn't be an issue when used for image editing, but for image generation, you may need to adjust the prompt with more detailed input to achieve compositional variety.
Lines at the Image Edges
In MMDiT-generation image generation models like Qwen-Image and Flux.1, lines can occasionally appear at the edges of the image when an illustration is generated.
In Qwen-Image-Edit-2509_clear, these lines are more noticeable because the output contrast has been increased.
Currently, there is no fundamental way to eliminate these lines. If they are a concern, you can use nodes like Image Inset Crop (rgthree) from rgthree-comfy custom nodes to crop the edges.
Can LoRA Be Used?
Since the Qwen-Image model is released under a free license, a large number of community-developed LoRAs exist.
While we haven't specifically tested individual LoRAs, the Qwen-Image-Edit-2509_clear model is only a fine-tuned version of Qwen-Image-Edit-2509.
Therefore, I believe that LoRAs designed for Qwen-Image-Edit-2509 should still be effective.
Workflow
Finally, here are the ComfyUI workflows for using Qwen-Image-Edit-2509_clear.
Text-to-Image
Image-To-Image
Models
Conclusion: Give Qwen-Image-Edit-2509_clear a Try
- Qwen-Image-Edit-2509_clear is now available.
- Superior texture and vivid contrast.
- Completely free commercial use.
In this article, I introduced the Qwen-Image-Edit-2509_clear model.
While local AI image generation models are smaller and generally slower than online models, the ability to pursue the best possible settings gives them the potential to surpass the expressiveness of online models.
Qwen-Image-Edit-2509_clear is released under a license that is completely free for commercial use, so I encourage everyone to give it a try.
Thank you for reading to the end!