Google levels up AI imaging with powerful new Nano Banana Pro

Google is expanding its generative AI lineup with the introduction of Nano Banana Pro. Built on the original Nano Banana’s viral popularity, the new version is available through the Gemini mobile app. It runs on the recently released Gemini 3 Pro architecture, designed to manage more advanced, intricate image generation and editing tasks than earlier systems.

Nano Banana Pro aims to address challenges that previous AI image generators frequently faced, offering more consistent accuracy in both generating and modifying images, even with detailed or complex prompts. Earlier models like Gemini 2.5 Flash Image sometimes struggled with precise instructions, but the updated system converts nuanced text prompts into visuals with coherent styling and clear, natural-looking text. This upgrade eliminates issues such as the “AI squiggles” that often appeared in place of actual writing.

In contrast to earlier consumer-focused tools, which often produced results that lacked professional reliability, Nano Banana Pro introduces stronger reasoning and better contextual awareness. Users can combine up to 14 reference images, and the model can consistently preserve the distinct characteristics of up to 5 people, enabling more reliable group composites. The system also improves stability in repeated outputs, keeping key traits such as facial structure, color schemes, and clothing consistent across multiple generated images within the same session.

Another significant upgrade allows users to upload photos or previously generated images for precise refinement. They can modify specific elements such as lighting, color grading, or camera angle while preserving the rest of the composition. This selective editing helps prevent unexpected visual artifacts, a common issue with less capable models.

The new model also supports image generation at resolutions up to 4K, making it suitable for commercial uses such as marketing assets, data graphics, and complete infographic designs, all producible in a single step.

Every image produced by Nano Banana Pro contains embedded SynthID watermarks, which can be detected through Google’s verification tools. This invisible watermarking, coupled with expanded C2PA metadata, plays a key role in transparency efforts by confirming when content is AI-generated, especially as visuals become increasingly realistic and manual verification grows more difficult. The Gemini app now allows users to check uploaded images for AI provenance and identify whether they originate from Google’s system by scanning for SynthID.

Subscribers on the AI Ultra tier can generate images without the visible watermark overlay, though the invisible SynthID marker remains for verification. Free users and those on lower tiers will continue to see the Gemini twinkle watermark displayed in the corner of generated images.