Observe ZDNET: Add us as a most popular supply on Google.
ZDNET’s key takeaways
- The brand new Nano Banana 2 guarantees higher picture technology.
- The brand new mannequin goals to supply high-quality photographs a lot quicker.
- Nano Banana 2 is now the default mannequin throughout all Google apps and websites.
Final yr Google made an enormous splash on the planet of AI-powered picture creation with its Nano Banana mannequin. Now the subsequent technology has debuted. And it guarantees a bunch of enhancements for producing pictures, drawings, diagrams, infographics, and different photographs for people, professionals, and companies alike.
On Thursday, Google introduced Nano Banana 2 (formally dubbed Gemini 3.1 Flash). Accessible to free and paid customers, the brand new mannequin replaces the primary Nano Banana in all places. Which means it is now the default on the Gemini web site and app, Google search, Google Move, AI Studio, and the Gemini API for builders.
Additionally: The most effective AI picture mills of 2026: There’s just one clear winner now
So what makes the second banana so significantly better than the primary? Here is what’s new, in keeping with Google.
1. Superior world data. Nano Banana 2 makes use of Gemini’s real-world data and pictures from net searches to attempt to extra precisely create your photographs. Google stated that this ability will assist the AI higher devise infographics, flip your notes into diagrams, and visualize your knowledge
2. Consistency. One flaw with AI picture technology has been an absence of consistency throughout a number of photographs. Ask the AI to regenerate a picture or create a brand new one primarily based on a earlier picture, and the characters and different components do not all the time stay the identical. Nano Banana 2 goals to be sure that as many as 5 characters and as much as 14 objects in a picture keep constant by means of every iteration.
3. Higher textual content rendering. Producing textual content in an AI picture may also be hit and miss, with illegible or inconsistent outcomes. Nano Banana 2 will attempt to create extra correct and legible textual content, serving to it design such photographs as greeting playing cards and mockups for promoting campaigns. The brand new mannequin may even translate textual content in a picture, so you possibly can share variations of the identical one throughout completely different languages.
4. Higher at following your directions. One other weak spot with AI picture mills is that they do not all the time observe your directions appropriately, particularly in case your description is advanced. Nano Banana 2 is designed to keep away from this downside by capturing the tone and nuances of your immediate, so you need to get the picture you need.
Additionally: Tips on how to spot an AI picture: 6 telltale indicators it is faux – and my go-to free detectors
5. Sooner speeds. With the earlier technology of bananas, you’d have to decide on between high quality and pace. Nano Banana Professional would generate a high-quality picture, however would possibly take some time. Nano Banana Flash would pace up the method, however the ensuing picture won’t be pretty much as good. The brand new model tries to mix the most effective of each worlds by producing higher-quality photographs at quicker speeds.
6. Larger constancy. With Nano Banana 2, Google stated that your photographs will characteristic sharper particulars, richer textures, and extra vibrant lighting. And the method runs extra rapidly than up to now.
7. Manufacturing-ready high quality: Seeking to create a brand, infographic, commercial, or different picture that should look skilled? With Nano Banana 2, you possibly can management the facet ratio of the picture to ensure it is sized for the suitable medium. To generate sharp photographs, it’s also possible to set the decision anyplace from 512px to 4K.
That every one sounds cool. However is the brand new Nano Banana all that Google guarantees? I attempted a few completely different requests to see the way it fared. These had been prompts that Gemini itself prompt to check the textual content rendering and the character consistency.
Additionally: I examined the brand new ChatGPT Photos – it is a gorgeous enchancment, and enormously enjoyable
For the primary one, I submitted the next immediate:
“A vintage-style journey poster for a lunar colony on the Moon. Within the heart, a smooth Nineteen Fifties-style rocket is touchdown close to a glass dome. On the high, in a daring, cream-colored Artwork Deco font, render the phrases ‘VISIT LUNAR CITY.’ On the backside, in smaller however clear script, add the phrase “The Gateway to the Stars.” Use a palette of deep navy, muted orange, and silver. Excessive decision, skilled graphic design model. Use a 16:9 facet ratio and create the picture in 1080p.”
The ensuing picture captured the suitable model and font for the textual content. The AI additionally adopted my directions for the colours, facet ratio, and determination.
For the second, I submitted two separate requests to check the character consistency from one picture to the subsequent.
The primary immediate learn:
“Create a personality sheet for ‘Captain Nova,’ a feminine house explorer with brief, vivid purple hair, a glowing blue mechanical eyepiece over her left eye, and a silver-and-white flight swimsuit with a gold solar emblem on the chest. Present her from the entrance, aspect, and again towards a plain gray background. Create the picture in a 21:9 facet ratio at 4K.”
Right here, the AI adopted my directions with a sheet that confirmed the character and highlighted a number of key particulars.
However the actual check was in my subsequent immediate, which learn:
“Utilizing the identical character of ‘Captain Nova,’ present her working by means of a neon-lit cyberpunk market. She is trying over her shoulder with a decided expression, her purple hair wind-blown. Maintain her mechanical eyepiece and the gold solar emblem on her swimsuit precisely the identical. Cinematic lighting, shallow depth of subject. Once more, hold the picture in a 21:9 facet ratio at 4K.”
This one stored the character constant in model and look, so she seemed the identical in each photographs.
Primarily based on these prompts, Nano Banana 2 looks like a worthwhile improve over the prior technology. That is excellent news since it’s now the default throughout the board. Google AI Professional and Extremely subscribers can nonetheless use the legacy Nano Banana Professional for extra specialised duties. In any other case, anybody can check out the newest taste of the brand new banana to see what photographs it unpeels.
