Google finally unveiled Nano Banana 2, the next version of its highly popular Nano Banana image model. This new model is based on the Gemini 3.1 Flash image model, so it's fast and also 50% cheaper than the pro model.

Nano Banana 2 is now leading the text-to-image ranking on Arena AI.

Here are some of the key technical features:

• Real-Time world knowledge: Integrated Google Search and Image grounding for accurate rendering of real-world landmarks and facts.
• Multi-subject consistency: Industry-leading identity preservation for up to 4 characters and 10 objects across a sequence.
• Advanced typography: Precise text rendering in multiple languages, perfect for professional infographics and global ad campaigns.
• Chain-of-thought reasoning: Built-in "Thinking Level" configuration that analyzes complex prompts for superior logic and instruction following.
• 4K ultra-HD & extreme ratios: Native 4096×4096 resolution with support for specialized formats like 1:8, 8:1, 1:4, and 4:1.

Aside from the features above, Nano Banana 2 also supports up to 14 image inputs, generates up to 4K image outputs, and can handle complex spatial instructions.

Now, let's talk about where to access Nano Banana 2.

Google has made the model accessible across their ecosystem. You can try it on Gemini, AI Studio, Vertex AI, Google Flow, and even on Google Ads.

Most of the regular users would use the Gemini app, but there's one big catch. There are not many editing controls that you can toggle to customize the output. Take note that Nano Banana 2 supports more aspect ratios and multiple resolutions. Another thing is the AI watermark embedded in the lower right corner of the image, which is kinda annoying if you ask me.

So, what I usually suggest using is a platform designed to expose these creative controls with fast inference and low cost. One such platform is Pollo AI.

Take note that Pollo AI is currently offered at 80% off. If you are interested in exploring this new model, I highly recommend taking advantage of this limited campaign promo.

Head over to Pollo AI's website and open the AI image tab. This is what you'll see on the dashboard:

From here, describe the image you wish to generate and adjust the customization controls like the aspect ratio, the resolution, and the number of output images.

Here's an example:

Prompt: Image of a young asian model doing a street photoshoot

Here's another one that showcases the photorealism capability of the model:

Prompt: Cinematic still in the style of a surreal dream sequence with a full shot of a woman standing in a field. The camera captures her from a slightly low angle, emphasizing the dramatic flow and volume of her fantastical garment. The dress is a multi-layered creation of sheer fabrics in pastel shades of lime green, pale pink, and soft yellow, billowing around her as if caught in a gentle breeze, creating abstract shapes and a sense of weightlessness. The woman has short, light pink-toned hair and a delicate expression. The background consists of a vast, golden grassland stretching towards distant, softly rendered mountains under a clear, light sky. The lighting is soft and diffused, contributing to the ethereal and dreamlike quality of the scene.

Each image is done in less than 10 seconds of inference time. It's not the fastest model out there, but this is still incredibly fast considering how good the quality of the image is. This is perfect for companies looking at solutions that can do batch image generation without compromising the quality.

Another impressive feature of Nano Banana 2 is its real-time world knowledge and grounding. Unlike other models that "hallucinate" details, it can research real-world architecture, historical facts, or scientific data before rendering to ensure visual accuracy.

Check out this example:

Prompt: Infographic about the water cycle

You can bet that the information presented in this image is correct because the image model connects directly to Google's live knowledge base.

One final thing to note is Nano Banana 2's improved capability of rendering texts more accurately. Here's an example:

Prompt: A cinematic close-up of a wooden sign made of recycled materials. The sign features detailed illustrations of local birds and the text "Native Wildlife: Please Observe from a Distance" in elegant, legible handwriting.

It can render sharp, legible fonts and can even translate your creative concepts into different languages while adjusting the visual atmosphere to match the target culture.

Fina Thoughts

Nano Banana 2 isn't the fastest model right now, and it's also not the cheapest. But the quality of the results you get kind of justifies the speed and cost. My favorite new features of this model are the improved typography and the strict prompt adherence. I usually use image generators for infographics and AI article cover images, so this might be my new go-to image model.

It is perfectly suited for applications powered by APIs, real-time products for users, and large-scale content needs like e-commerce catalogs or quick prototyping, where maintaining a low cost per image is critical.

So, as I already mentioned, Pollo AI is offering access to Nano Banana 2 at an 80% discount right now. It's a great opportunity to explore the model while the promo lasts. Let me know your thoughts in the comments!

This story is published on Generative AI. Connect with us on LinkedIn and follow Zeniteq to stay in the loop with the latest AI stories.

Subscribe to our newsletter and YouTube channel to stay updated with the latest news and updates on generative AI. Let's shape the future of AI together!