This Mystery AI Called 'Nano Banana' Is Beating Every Image Editor on LMArena

Google has launched a major upgrade to Gemini's image editing capabilities with the introduction of its new "Nano Banana" image editing model from Google DeepMind, which has already achieved the top rating as the world's best image editing model.
 

The new model excels at maintaining character consistency across edits.

When editing photos of yourself, family, or pets, the system ensures subjects look consistently like themselves, whether you're trying a 1960s hairstyle or putting a tutu on your pet.

Advanced Editing Capabilities Include:

• Photo Blending: Upload multiple photos and blend them together for brand-new scenes, like creating a portrait of you and your dog on a basketball court.
• Multi-turn Editing: Continuously edit images Gemini creates - paint walls in an empty room, then add furniture piece by piece.
• Style Transfer: Apply textures and colors from one image to objects in another, like using flower petal patterns for dress designs.

Test the Model Yourself on LMArena:

Both paid and unpaid users globally can access this updated image editing capability in the Gemini app starting today. You can also test and compare the model's performance at lmarena.ai.

How LMArena Works:

LMArena is a crowdsourced AI evaluation platform where you're served two anonymous models that respond to your prompt, and you vote on which response is better. 

Model identities remain anonymous until you've voted. Sometimes one of the anonymous models in LMArena's battles is Google's Nano Banana model, allowing you to test it alongside other leading AI image editors without knowing which is which until after you make your choice.

With over 3.5 million votes from the global community, LMArena has become the de facto standard for comparing AI models, making it the perfect place to experience Google's latest breakthrough in image editing technology.

All images created or edited include visible watermarks and Google's SynthID digital watermark to clearly identify AI-generated content.