Google has rolled out a major upgrade to its Gemini AI chatbot’s photo editor, powered by a new DeepMind image model nicknamed “Nano Banana” . Dubbed Gemini 2.5 Flash Image in Google’s developer documentation, this state-of-the-art model can blend and edit photos with unprecedented fidelity . It’s now live in the Gemini app for all users (free and paid) and available via the Gemini API and Google AI Studio for developers . According to Google, early testers are already “going bananas” over it – the company claims it’s the top-rated image editing model in the world .
Key Features of Nano Banana
- Character consistency and transformations: The editor preserves a person’s or pet’s appearance even after big changes, so you can swap hairstyles or outfits without losing identity . (Imagine reimagining yourself with a 60’s beehive haircut or in a matador costume – the result still looks like you.) .
- Photo blending: You can upload multiple images and merge them into one scene. For example, Gemini can combine a portrait and a pet photo into a single image of you and your dog together on a basketball court .
- Multi-turn editing: Gemini now supports iterative, step-by-step edits. Start with a base image (say, an empty room), then ask Gemini to paint the walls, next to add furniture, next to place a coffee table, etc. Each new edit only changes what you specify, keeping the rest of the image intact .
- Style transfer (design mixing): The model can apply the style or pattern from one image to another object. For example, you could take the colors and texture of a flower photo and use them to design a dress or even restyle a pair of rainboots .
- Precise prompt-based edits: In addition to big-picture changes, you can make fine-grained edits via natural language. Nano Banana can blur a background, remove stains or objects, erase people from a photo, tweak a subject’s pose, or colorize a black-and-white image – all with a simple text instruction .
Character Consistency and Transformations
One standout trick is how Nano Banana keeps subjects looking like themselves through wild transformations. You just feed Gemini a photo and tell it what to change. The AI might turn you into an artist with a paintbrush or a 90s sitcom character with big hair — and the person still looks like the original . This “character consistency” focus means that faces, body shapes and even a pet’s fur pattern are preserved across edits, avoiding the distorted results that older AI tools sometimes produce . In Google’s words, you can put someone “anywhere in the world you can imagine — all while keeping you, you” . In practice, this makes it easy to experiment with outfit or environment changes (say, putting yourself on the moon or in a historical costume) without losing realistic detail.
Photo Blending and Iterative Edits
Nano Banana also excels at composite scenes. For example, upload one photo of yourself and another of your dog, then ask Gemini to “blend” them into a single image – say, you cuddling the dog on a basketball court. The model intelligently merges the subjects, background and lighting into a convincing new photo . Beyond single-step blends, Gemini now lets you iterate on an image in multiple turns. You might start with an empty living-room photo, tell Gemini to paint the walls blue, then instruct it to add a sofa, then a bookshelf. With each step the model updates just the requested parts and “remixes” the scene, leaving the rest untouched . This mimics a real design workflow, only your commands are simple text prompts.
Style Transfer and Fine-Grained Edits
The new model also brings creative style-mixing into your hands. For instance, you can take the intricate pattern of a butterfly’s wing and reimagine that pattern on a dress. In tests, Gemini painted a street scene model’s gown with the bright blue-and-orange butterfly motif – a striking, photorealistic result . On top of these artistic effects, Nano Banana can handle surgical edits as well: you can tell it to blur the background, remove an unwanted object or even an entire person, tweak a subject’s pose, or colorize a monochrome photo with just a sentence . These precise edits work through natural-language instructions, turning complex Photoshop-like tasks into simple chat commands.
What This Means for Image Workflows
Google says Nano Banana leverages Gemini’s built-in world knowledge to make context-aware edits , and indeed benchmarks list it as state-of-the-art. For tech teams and creatives, that means much faster prototyping and editing. Designers or marketers can now mock up visuals, iterate ad concepts or style assets by talking to Gemini, rather than manually slicing and editing images. All Gemini-generated or edited images carry a visible “AI” watermark and an invisible SynthID code to flag them as AI-created , ensuring clear attribution. And because the same model is available via the Gemini API and Google AI Studio , companies can plug Nano Banana into their own tools (for example, automating product mockups or virtual staging). In short, Google’s Nano Banana upgrade substantially expands Gemini’s image editing capabilities – combining higher-quality results with more intuitive, multi-step control, and paving the way for new AI-powered imaging workflows.
