Google has unveiled an updated image generation model, Gemini 3.1 Flash Image, codenamed Nano Banana 2. It will replace Nano Banana Pro in the Gemini app, but users of Google AI Pro and Ultra subscriptions will retain access to the Pro version.

The new model will be available in AI Mode and Lens in Google Search, in AI Studio and through the Gemini API in preview mode, on the Google Antigravity platform, in Vertex AI, and will also become the default model in the Flow video generation service.

Key improvements include the use of the Gemini knowledge base and search data to more accurately generate specific objects, as well as improved text processing: the model can accurately generate and translate captions on images.

Compared to the previous version, the model maintains better consistency: the appearance of up to five characters and the appearance of up to 14 objects remains unchanged within a single project. This allows for the creation of source footage for videos with precise detail.

The model also more accurately follows user instructions and handles complex queries. It supports aspect ratios and resolutions from 512px to 4K.