- Google’s Gemini Nano Banana 2 AI model is expected to go live soon.
- It will be based on the Gemini 3.0 Pro and use a new multistep generation workflow
- Here’s how the image generation model will work.
Earlier this year, Google unveiled its Gemini Flash 2.5 model, called Nano Banana. The image/video generation model took the world by storm for its realistic, detailed outputs. Google is reportedly working on its successor, aka Nano Banana 2, and it’s just around the corner. Ahead of the official launch, images captured by the model have gone viral, showcasing its capabilities. Here’s what you can expect with Google’s Nano Banana 2 model.
Gemini Nano Banana 2 Generates Images With Complex Text
According to a report by Testing Catalog, the image generation model was available briefly for the Media AI platform, hinting at its imminent release. While the initial version is based on the current Gemini 2.5 Flash model, it is likely to transition to Gemini 3.0 Pro in the future.
The report further claims that Nano Banana 2 uses a new multi-step generation workflow. This means it first analyzes the user’s requirements, spends a lot of time planning the output, and then generates an image.
Also Read:
- Unlocking Google’s Secret: How to Use Nano Banana in Google Lens and NoteBookLM
- How I Create Storybooks With Illustrations Using Gemini
Nano Banana 2 also uses a built-in image analyzer to review the output and rectify any errors. This multi-step process is currently missing in Nano Banana 1.0, and it should help improve the accuracy of its successor.
It can also come in handy for detailed and complex images where accuracy is the key. Other expected changes include support for various aspect ratios such as 1:1, 2:3, 3:2,3:4, 4:3, 9:16, and 21:9.
Users can expect several changes with Nano Banana 2, including crisper images at 2K with the option to upsample to 4K. Leaks suggest there will be “significant improvements” when it comes to text rendering, charts, infographics, world knowledge, etc.
Nano Banana 2 Is Already Taking Over Social Media
Several users on X have shared output by Nano Banana 2, internally called GEMPIX 2. These images showcase the wide range of image generation capabilities of the model. From generating images of anime like Tokyo Ghoul to robots working in a cyberpunk dystopian environment, the output is a huge upgrade in terms of visuals.
Some outputs even showcase the text generation prowess of Nano Banana 2 in images, something which many chatbots like ChatGPT fumble at. We can see Nano Banana excelling at complex tasks, such as generating images of UI, such as Windows 11.
In one example, a user managed to turn the water pink using Nano Banana 2, a scenario where its predecessor failed. Another instance showcased the AI model upscaling GTA San Andreas into a faithful masterpiece.
There’s no doubt that Google’s Nano Banana 2 model will be a massive success once it rolls out commercially. The AI model is expected to roll out from November 11; however, we have yet to see any official news at the time of writing this article.
