Premium

Nano Banana 2: Google’s next AI model promises smarter image understanding

The upcoming Google Nano Banana 2 will get a better context window and better resolution, and a new aspect ratio.

According to reports, the model will evaluate its own output, spot discrepancies, and autonomously make corrections, increasing accuracy and realism over earlier iterations. (Image: Google)

Google has unveiled its Gemini 2.5 Flash model, known as Nano Banana, which has recently attracted attention for its sophisticated image-editing and natural language image-generating features. Social media users have begun posting samples of its enhanced features.

Sources claim that the next Nano Banana 2 model will excel in text rendering, infographics, charts, and user instruction compliance. It might provide a greater variety of aspect ratios and output resolutions.

According to reports, the Gemini 3.0 model is set to be launched soon, which may improve the performance of the future Nano Banana 2.

What to expect?

According to reports, that Nano Banana 2 will employ a novel multi-step generating process that plans, evaluates, and improves outputs before displaying the finished image.

Story continues below this ad

Also Read | From hands-free navigation to accident-prone area alerts: Here are the top features coming to Google Maps in India

According to reports, the model will evaluate its own output, spot discrepancies, and autonomously make corrections, increasing accuracy and realism over earlier iterations. More aspect ratios (1:1, 2:3, 3:2, 4:3, 16:9, 21:9) and output resolutions (1K to 4K) are reportedly supported by the upgrade.

The Gemini 3 Pro Image model, which powers the Nano Banana 2, is expected to significantly enhance fidelity, particularly for fine text and visual details, according to TechRadar’s research. This implies that posters and other design graphics produced by AI might have crisp lettering and crisp edges.

Additionally, the model is expected to enable native 2K renderings with 4K upscaling, designed for professional designers who require outputs ready for export.

Improved comprehension, character consistency and editing software

Nano Banana 2 is also expected to improve cultural context awareness. Prompts such as “a family picnic in Tokyo springtime” or “streetwear shoot in Berlin winter” will reportedly generate more geographically accurate visuals. The model is reportedly being trained on broader regional datasets to make scenes feel more authentic.

Story continues below this ad

The Nano Banana tool’s character consistency is already one of its main features; nevertheless, the enhanced version is supposed to enable smoother consistency when creating a sequence of photographs with the same subject or person.

Google is also reportedly testing a new feature called Edit with Gemini, which would allow users to edit portions of existing photos without having to start from scratch. For instance, users might utilise a straightforward highlight-and-edit procedure to change backgrounds, clothing, or lighting.

Tags:
Google