All of us ought to admit that September has been the month of Gemini Nano Banana. Everybody has been enjoying round with the mannequin, posting quirky edits and enjoyable selfies on social media. Likelihood is, you tried it too or no less than scrolled previous a dozen of these movie star mashups and 3D figurine posts in your feed. However whereas Nano Banana has dominated the informal facet of picture era, Qwen has been quietly leveling up the skilled modifying recreation. Simply final month, the crew launched Qwen-Picture-Edit, and now they’re again with an upgraded model: Qwen-Picture-Edit-2509, that focuses on consistency, flexibility, and management. In comparison with the August launch, the brand new replace makes modifying extra highly effective for creators, builders, and researchers, whereas nonetheless staying accessible to anybody who needs to experiment.
If you wish to know extra in regards to the earlier mannequin, learn our detailed article on Qwen-Picture-Edit!
What’s New in Qwen-Picture-Edit-2509?
In comparison with final month’s launch, this model introduces 4 main enhancements:
Multi-image Modifying Assist
Qwen-Picture-Edit-2509 now means that you can edit throughout a number of pictures concurrently. The mannequin has been educated utilizing picture concatenation, enabling seamless edits throughout combos like:
- particular person + particular person
- particular person + product
- particular person + scene
It really works finest with 1–3 enter pictures, and in addition integrates effectively with ControlNet maps akin to keypoints for pose adjustments.
Enhanced Single-image Consistency
When modifying a single picture, the September replace focuses closely on consistency:
- Particular person Modifying: Faces retain their id throughout completely different poses and portrait types.
- Product Modifying: Logos and objects preserve their look, making product posters simpler to generate.
- Textual content Modifying: Not solely can you modify the phrases, however you can even alter fonts, colours, and supplies with precision.
Native ControlNet Assist
The replace provides easy integration with ControlNet inputs like depth maps, edge maps, and keypoint maps, increasing artistic management for technical and inventive use instances.
Improved Textual content + Picture Integration
Qwen-Picture-Edit-2509 can now mix textual content modifying seamlessly with picture manipulation. For instance, designing a poster the place fonts, types, and visuals work together constantly.
Additionally Learn: Qwen3-Omni Overview: Multimodal Powerhouse or Overhyped Promise?
Arms On Prompts to Attempt with Qwen-Picture-Edit-2509
For all these duties, I’m utilizing each Qwen-Picture-Edit-2509 and Nano Banana to match the outputs and in addition see how Qwen performs on every job.
Job 1: Particular person + Particular person
Immediate: “Use picture A and picture B. Merge them into one picture the place each folks sit on a park bench. Protect face id and pure lighting.”
The brand new characteristic of the Qwen Picture Edit mannequin nonetheless struggles to maintain facial options intact. As you’ll be able to see within the picture above, Nano Banana did a greater job of retaining the faces of each folks. Qwen additionally altered the bag within the unique picture, whereas Nano Banana stored it unchanged.
Job 2: Particular person + Scene
Immediate: “Place this portrait right into a busy Tokyo road at evening with neon indicators. Maintain the face id and proper pores and skin tone.”
Qwen did an important job in retaining the facial expressions and general picture high quality. The lightning and general picture mix seems to be good. Nonetheless, Nano Banana went one step forward and edited the background lightning as per the picture inflicting it to look extra actual.
Job 3: Particular person + Product Advert
Immediate: “Use this portrait and this product picture. Make it appear to be an advert the place the particular person holds the product. Maintain product label readable.”
For this job, I added the pictures of each the mannequin and the product into each fashions utilizing the identical immediate. The output from Qwen seems to be higher, whereas Nano Banana merely positioned the 2 images collectively with out a lot thought in combining them.
Job 4: Pose Management with Keypoint Map
The next pictures reveals the method the place a keypoint pose map from one picture is used to switch a particular posture to an individual in a distinct picture, all whereas preserving their facial id. The examples additionally present how this know-how may be mixed with different advanced modifying duties, akin to altering the background and manipulating objects.
I’ve used this new characteristic of Qwen within the following immediate:
Immediate: “Use this portrait and this keypoint pose map. Repose the particular person to match the map whereas maintaining facial id.”
Qwen Picture Edit has specific help for keypoint-guided modifying. That’s why your sketch + keypoint overlay synced completely. It is aware of how one can learn the keypoint map, then situation the era. If you wish to play with poses, angles, posture, or interactions with landmarks (like Eiffel Tower, Qutub Minar, Taj Mahal, Akshardham), it’s best to stick to Qwen Picture Edit (or any mannequin that mentions “pose management” / “keypoint conditioning”).
Nonetheless, Nano Banana doesn’t have a keypoint-map conditioning pipeline. That’s why it simply re-emitted your enter picture with out adjustments: it could edit for model, texture, background swaps, and many others., but it surely doesn’t parse or implement human/object pose maps.
Job 5: Including Textual content to Picture
Immediate: “The woman within the uploaded picture is holding a chalk board with the textual content “Be part of my GenAI Masterclass to Grow to be an Skilled in 2025″ written on it”
Between the 2 responses, Qwen’s output is clearly higher. Within the Qwen model, the textual content sits naturally on the board, aligned correctly with the attitude, and appears prefer it really belongs on a chalkboard. The font model is constant, clear, and extremely readable, which makes the message come throughout extra clearly. The textual content blends seamlessly with the board floor, so the ultimate end result feels genuine and reasonable.
However, Gemini’s response seems to be extra like a digital overlay. The textual content alignment doesn’t absolutely match the angle of the board, and the font model seems much less like chalk writing and extra like commonplace digital textual content. Due to this, the textual content seems like it’s pasted on high of the board moderately than built-in into it.
Job 6: Cartoon or Anime Variant
Immediate: “Flip this image into an anime model character whereas preserving the identical face id and expression.”
Between the 2 outcomes, Qwen-Picture-Edit-2509’s model is clearly higher as a result of it absolutely transforms the particular person into an anime-style character whereas retaining the identical outfit, pose, and expression, making it devoted to the duty immediate, whereas Nano Banana’s model seems to be nearly equivalent to the enter picture with little or no anime-stylization, so it doesn’t actually fulfill the “anime character” requirement and feels extra like a light-weight retouch than a metamorphosis.
Job 7: Product Poster
Immediate: “Make a clear marketable poster for this shoe. Use a plain background. Add the product identify on the backside in daring kind saying – SoulShoe”
Between the 2, Qwen’s poster is healthier as a result of it goes past a plain showcase and really markets the product. The glowing impact makes the footwear stand out, the tagline “The place Consolation Meets Fashion” provides emotional enchantment, and the “Restricted Version” badge creates urgency, all of which make it appear to be a sophisticated advert marketing campaign, whereas Gemini’s output, although clear {and professional}, feels extra like a easy catalog picture than a marketable poster.
My Verdict
Nano Banana is enjoyable, accessible, and immediately shareable; good for memes, selfies, and social tendencies. Qwen, alternatively, is structured, versatile, and production-ready, making it the higher selection for creators, designers, and professionals who need reliability in outputs.
Additionally Learn:
Conclusion
AI picture fashions are bettering month by month. August gave us the primary Qwen Picture. September has given us this new model with multi picture help and higher consistency. Subsequent month who is aware of what is going to come. In the event you love experimenting then open Qwen Chat and check out the prompts I shared. Begin with one thing easy like inserting your self in a brand new background. Then transfer to multi picture edits and posters.
I’ll hold testing and sharing what works. In the event you strive any of those prompts let me know what end result you bought. Typically the enjoyable is within the surprises.
Hiya, I’m Nitika, a tech-savvy Content material Creator and Marketer. Creativity and studying new issues come naturally to me. I’ve experience in creating result-driven content material methods. I’m effectively versed in website positioning Administration, Key phrase Operations, Net Content material Writing, Communication, Content material Technique, Modifying, and Writing.
Login to proceed studying and revel in expert-curated content material.
