Google Gemini's new Nano Banana Pro is breaking the internet

Published on

Another ABSURD piece of news from Google/Google DeepMind that I’ve had access to for a while and was itching to talk about! Today they launched “Nano Banana Pro”, their new AI image generation/editing model, based on Gemini 3 Pro.

Speaking of text, it looks like we’ve finally solved that old problem of AI generating hieroglyphs. The original Nano Banana was already great, but it wasn’t perfect (just like GPT-4o isn’t either). From my tests so far, I’ve (rarely) seen a tiny mistake here and there (when there’s A LOT of text in the same image), but overall, it writes perfectly, and it also gives you the freedom to choose fonts and styles.

A huge problem with the original Nano Banana that they fixed is that, now, this new version respects your format request (16:9, 1:1, 9:16, etc). That’s a huge deal.

Another REALLY cool thing is that it now has access to real-time internet data via Search Grounding, so you can ask it to create an up-to-date infographic about the weather, or a visual step-by-step of a recipe, and it generates everything with the correct, current information.

According to the announcement, it can keep the identity of up to 5 different characters and accepts up to 14 reference images to compose a scene. For anyone who creates storyboards or ad campaigns, this is the “holy grail” that was missing.

The model is already being rolled out today in Gemini for everyone (by selecting the “Thinking” mode).

I ran a few tests here (which left me open-mouthed), and below I’m leaving the prompts and the resulting images (the prompts were originally written in Portuguese):

1 - “Realistic image of a cutting board with several fruits being cut. The phrase ‘Nano Banana Pro’ is formed from pieces of banana, shown larger. Below it, smaller, the phrase ‘Gemini 3 Pro’ formed from pieces of other fruits (strawberries, apples, oranges, kiwis, etc).”

Image generated by Nano Banana Pro of cut fruits

2 - “I want you to search online for the top 5 phones sold in Brazil for less than 3000 reais. Then, create an infographic with the pros and cons of each one.”

Image generated by Nano Banana Pro of the top 5 phones

3 - “I’m sending you attached the drawing of the Transformers architecture from the paper ‘Attention is All you Need’. I want you to recreate the image with annotations in English next to each element, explaining what each of them does, and also how they work together to make an LLM work.”

Image generated by Nano Banana Pro of the Transformers architecture

4 - “This is a photo of me in South Korea. I want you to remove the people in the background of the image and make the day sunnier. Also, swap my clothes for the t-shirt and the visor in the images I sent you, keeping the image format at 9:16.”

Image generated by Nano Banana Pro of me in South Korea

5 - “Create an infographic with annotations explaining how the MoE (Mixture-of-Experts) architecture works in LLMs, 16:9 format.”

Image generated by Nano Banana Pro of an infographic

Obviously, we’ll comment on it in more depth in the next news episode of the IA Sob Controle podcast 😉