Just now, the moment it opened Gemini the page, it will pop up Google Newest and most powerful Nano Banana Pro Update notice: this update is really impressive, directly improving what came before. Nano Banana The image resolution is low, with a bunch of garbled Chinese characters and various defects. Ada gives everyone a quick rundown of the new features of the Nano Banana Pro.
Google launches more powerful Nano Banana Pro image generation model
1. Generate text:
Previously, with any image generation AI on the market, Chinese characters were almost always their Achilles’ heel. This time, one of the biggest improvements Nano Banana Pro brings is that it can now display Chinese characters correctly! Take the following prompt as an example:
The generated banner has completely accurate Chinese text, and you can specify the style. It’s also unaffected by the dimensions of the uploaded image, allowing you to set the size.
Generating posters is no problem either—just a simple sentence description, the specified image ratio, and the text to be displayed.
If you write the prompt carefully, the performance will be better:
There’s no such event, everyone, don’t go, hey:
Second, better semantic understanding and better quality.
The previous Nano Banana was already great at drawing, but this version takes generation quality to the next level, with highly accurate capture of characters’ essence.
Understanding of instructions has also improved—asking it to generate a MAC screen browser and MKBHD YouTube channel view was rendered perfectly, even including the video thumbnails. The images also changed from the original 1000PX width to a default of 1408PX.
However, in Gemini, when I asked for a higher resolution it ignored me (the original image below is 1376*768). I think the character three-view quality is better than the previous generation. For higher resolution, you need to go to AI Studio, which supports up to 4K quality:
3. Maximum of six images per combination.
Previously, Nano Banana could also merge multiple images, but this time it’s even more impressive—it can combine up to six images at once. In the example below, even though I made typos and mixed up the order, the AI still understood what I meant.
Four, can draw short comics with story dialogue.
Now you can generate a short comic just by inputting photos and a description. A longer one should be feasible too, but it would take more trial and error. Dialogue can also be directly translated or generated in Chinese.
At the beginning, without specifying that it should be presented in Chinese, an English comic was produced:
Then I asked the AI to translate it directly into Chinese, and it just went ahead and translated it. The font wasn’t messed up either. Amazing.
5. Illustrated Explanatory Images
This is a step-by-step diagram of the government’s universal NT$10,000 payout, made based on prompts from netizens. I only typed in these words—everything else was generated entirely by AI. It’s incredibly powerful.However, it should be noted that such illustrated instructions may contain errors, so the content needs to be checked.）：
There are some other features Ada is still testing, so go ahead and try them out! Just switch the model to the “3 Pro thinking model” and select “create image” when giving a command. If you haven’t upgraded to Pro yet, just wait — it should be available soon:
Free users can make up to four images per day in Gemini; when you exceed that, wait until tomorrow (but you can still use the first-generation Nano Banana to continue making images). If you want to make more images, buy Pro:
Additionally, AI Studio has now been updated to the latest version of Nano Banana Pro (gemini 3 pro image Preview), with support for up to 4K resolution (available to paid users only).
If you feel that being able to make only four images a day is too few, you can also refer to this tutorial:
Source: KOCPC Chinese