Follow ZDNET: Add us as a preferred source on Google.
ZDNET's key takeaways
- The new Nano Banana 2 promises better image generation.
- The new model aims to produce high-quality images much faster.
- Nano Banana 2 is now the default model across all Google apps and sites.
Last year Google made a big splash in the world of AI-powered image creation with its Nano Banana model. Now the next generation has debuted. And it promises a host of enhancements for generating photos, drawings, diagrams, infographics, and other images for individuals, professionals, and businesses alike.
On Thursday, Google announced Nano Banana 2 (officially dubbed Gemini 3.1 Flash). Accessible to free and paid users, the new model replaces the first Nano Banana everywhere. That means it's now the default at the Gemini website and app, Google search, Google Flow, AI Studio, and the Gemini API for developers.
Also: The best AI image generators of 2026: There's only one clear winner now
So what makes the second banana so much better than the first? Here's what's new, according to Google.
1. Advanced world knowledge. Nano Banana 2 uses Gemini's real-world knowledge and images from web searches to try to more accurately create your images. Google said that this skill will help the AI better devise infographics, turn your notes into diagrams, and visualize your data
2. Consistency. One flaw with AI image generation has been a lack of consistency across multiple images. Ask the AI to regenerate an image or create a new one based on a previous image, and the characters and other elements don't always remain the same. Nano Banana 2 aims to make sure that as many as five characters and up to 14 objects in an image stay consistent through each iteration.
3. Better text rendering. Generating text in an AI image can also be hit or miss, with illegible or inconsistent results. Nano Banana 2 will try to create more accurate and legible text, helping it design such images as greeting cards and mockups for advertising campaigns. The new model can even translate text in an image, so you can share variations of the same one across different languages.
4. Better at following your instructions. Another weakness with AI image generators is that they don't always follow your instructions correctly, especially if your description is complex. Nano Banana 2 is designed to avoid this problem by capturing the tone and nuances of your prompt, so you should get the image you want.
Also: How to spot an AI image: 6 telltale signs it's fake – and my go-to free detectors
5. Faster speeds. With the previous generation of bananas, you'd have to choose between quality and speed. Nano Banana Pro would generate a high-quality image, but might take a while. Nano Banana Flash would speed up the process, but the resulting image might not be as good. The new version tries to combine the best of both worlds by generating higher-quality images at faster speeds.
6. Higher fidelity. With Nano Banana 2, Google said that your images will feature sharper details, richer textures, and more vibrant lighting. And the process runs more quickly than in the past.
7. Production-ready quality: Looking to create a logo, infographic, advertisement, or other image that needs to look professional? With Nano Banana 2, you can control the aspect ratio of the image to make sure it's sized for the right medium. To generate sharp images, you can also set the resolution anywhere from 512px to 4K.
That all sounds cool. But is the new Nano Banana all that Google promises? I tried a couple of different requests to see how it fared. These were prompts that Gemini itself suggested to test the text rendering and the character consistency.
Also: I tested the new ChatGPT Images – it's a stunning improvement, and enormously fun
For the first one, I submitted the following prompt:
“A vintage-style travel poster for a lunar colony on the Moon. In the center, a sleek 1950s-style rocket is landing near a glass dome. At the top, in a bold, cream-colored Art Deco font, render the words ‘VISIT LUNAR CITY.' At the bottom, in smaller but clear script, add the phrase “The Gateway to the Stars.” Use a palette of deep navy, muted orange, and silver. High resolution, professional graphic design style. Use a 16:9 aspect ratio and create the image in 1080p.”
The resulting image captured the right style and font for the text. The AI also followed my instructions for the colors, aspect ratio, and resolution.
For the second one, I submitted two separate requests to test the character consistency from one image to the next.
The first prompt read:
“Create a character sheet for ‘Captain Nova,' a female space explorer with short, bright purple hair, a glowing blue mechanical eyepiece over her left eye, and a silver-and-white flight suit with a gold sun emblem on the chest. Show her from the front, side, and back against a plain grey background. Create the image in a 21:9 aspect ratio at 4K.”
Here, the AI followed my instructions with a sheet that showed the character and highlighted several key details.
But the real test was in my next prompt, which read:
“Using the same character of ‘Captain Nova,' show her running through a neon-lit cyberpunk market. She is looking over her shoulder with a determined expression, her purple hair wind-blown. Keep her mechanical eyepiece and the gold sun emblem on her suit exactly the same. Cinematic lighting, shallow depth of field. Again, keep the image in a 21:9 aspect ratio at 4K.”
This one kept the character consistent in style and appearance, so she looked the same in both images.
Based on these prompts, Nano Banana 2 seems like a worthwhile upgrade over the prior generation. That's good news since it is now the default across the board. Google AI Pro and Ultra subscribers can still use the legacy Nano Banana Pro for more specialized tasks. Otherwise, anyone can try out the latest flavor of the new banana to see what images it unpeels.
