Nano Banana It has been very popular since its launch, not only because of its excellent generation quality but also because it maintains character consistency. As a result, there are now many different types of generation tools, such asProfessional portrait photo、A selfie with my childhood self、LINE sticker generation、Over 800 ways to play、Generate a three-panel snow scene photo.Wait. It’s been less than three months since Nano Banana launched, and I originally thought the next-gen model wouldn’t come until next year. But recently, a flood of Nano Banana 2 rumors have suddenly appeared, and just the other day, someone even AI The website was the first to use what appears to be a preview version of Nano Banana 2, and every generated AI imageAll quite stunning. The representative said Nano Banana 2 might be coming out soon.
Nano Banana 2 seems to be launching soon! Multiple viral generated image examples have surfaced
A few days ago, @testingcatalog announced on the X social platform that Google is preparing to launch Nano Banana 2, with the codename “GEMPIX2”, and it may be seen in a few weeks.
In addition, iLounge directly revealed in the article that Nano Banana 2, developed by Google DeepMind, is expected to be officially released in mid-November 2025, meaning we could see it as soon as next week. Rumors suggest Nano Banana 2 is based on the Gemini 3 Pro architecture and will push the limits of AI image generation even further, delivering unprecedented image quality and editing precision.
Earlier, @legit_api discovered that Media IO quietly released a preview version of Nano Banana 2, but it is recommended to remain cautious here, as Media IO’s reliability is yet to be confirmed:
Shortly after, many netizens tried using the Nano Banana 2 preview version to generate images, and here are the results compiled for your reference. As always, until Google officially announces Nano Banana 2, we can’t be 100% certain whether these images were actually generated using the new model.
Prompt: “Generate a screenshot of a Windows 11 desktop, with Google Chrome open, showing Mr. Beast thumbnails on YouTube.com.” (Source:@synthwavedd）
Generate an introduction page for Gemini 3 with a Windows 11 background:
Prompt: “The clock reads 11:15, and the wine glass is filled to the brim.” (Source:@synthwavedd）
Prompt: “A cyberpunk-style hacker robot working in front of multiple screens.” (Source:@testingcatalog）
Prompt: “Ken Kaneki embraces his friend in the snow, Tokyo Ghoul” (Source:@legit_api）
What improvements will the Nano Banana 2 bring? Rumors point to five key areas.
First is “sharper details and text integration.” Since Gemini 3 Pro Image is reported to significantly enhance prompt understanding and text generation capabilities, posters, UI, or magazine designs could all potentially render clear, readable fonts. It is expected to support native 2K output, with an upgrade option to 4K resolution.
Next is “cultural and geographical context awareness capability.” The model’s training data covers a broader range of geographical and cultural content to enhance authenticity across different regions. For example, when generating images such as “a family picnic in Tokyo in spring” or “a Berlin winter street fashion shoot,” it can provide more realistic local details.
“Character consistency has also been improved”—Nano Banana 2 once again strengthens “character coherence,” maintaining the same character’s appearance and actions across multiple scenes.
In addition to generating brand-new images, Nano Banana 2 will also support an “Edit with Gemini” mode. Users can directly select the area to modify, such as changing clothing, adjusting lighting, or replacing the background, without having to regenerate the entire image.
Finally, there’s “significantly improved generation speed.” Previously, generating an image might take 20–30 seconds, but the Nano Banana 2 is claimed to complete complex prompts within 10 seconds, putting its speed on par with Midjourney and Adobe Firefly.
Source: KOCPC Chinese