Google has announced the release of Nano Banana 2, officially named Gemini 3.1 Flash Image, its latest image generation model aimed at faster and more scalable visual creation. The new model follows the original Nano Banana, launched in August last year, and the Nano Banana Pro, which arrived in November.
Nano Banana 2 is positioned between speed and quality, combining advanced Pro-level visual reasoning with the low-latency performance of Google’s Flash architecture. According to Google, the update offers a significantly improved price-performance ratio for developers building image-heavy apps at scale.
Key features and improvements of Nano Banana 2
Nano Banana 2 introduces several improvements compared to its predecessor:
- Real-time knowledge integration: The model can reference real-world information and web search images to generate accurate visuals. Google demonstrated this with a demo of “Window Seat” that creates photorealistic views of windows based on real locations and live weather conditions.
- Improving the interpretation and localization of text: The template can generate readable, well-aligned text within images, enabling use cases such as user interface mockups and marketing creatives. It also supports localization on the image. A demo called “Global Ad Localizer” shows ad visuals automatically translated and adapted for different countries.
- Subject consistency across scenes: Nano Banana 2 supports visual consistency for up to five characters and 14 objects in a single workflow. Google demonstrated this through an example of “Pet Passport,” where a pet remains visually accurate to several global landmarks using a reference image.
- Configurable reasoning depth: Developers can check how much the model “thinks” before generating an image. Options include Minimal (default) and High/Dynamic, allowing for better tracking of instructions for complex requests.
- Extended output formats: Support has been added for extreme aspect ratios such as 4:1, 1:4, 8:1 and 1:8. A new 512px resolution level has also been introduced to reduce lag in high-volume pipelines, alongside the existing 1K, 2K and 4K outputs.
- Improved visual fidelity: Google says the model offers richer textures, more vivid lighting and improved detail while maintaining fast rendering speeds.
Google will continue to offer Nano Banana Pro for workflows that require the highest factual and visual accuracy, while Nano Banana 2 is optimized for rapid generation and search-based tasks.
Nano Banana 2 Availability
Nano Banana 2 is now rolling out to the following Google platforms:
- Developer Tools: Available via paid API access through Gemini API, Google AI Studio, Vertex AI, Google Antigravity and Firebase
- Gemini App: Replaces Nano Banana Pro as the default image template in Fast, Thinking and Pro modes; Pro and Ultra subscribers can still regenerate images using Nano Banana Pro
- Google search: Integrated AI mode and Google Lens on mobile, desktop and the Google app, expanding to 141 countries with support for eight additional languages
- Stream: Set as the default image generation pattern with zero credit usage
- Google Ads: Available for generating asset suggestions during campaign creation
Origin and verification of content
Images created by Nano Banana 2 include Google’s SynthID watermark, which works in conjunction with C2PA content credentials to provide transparency about how the content was created or modified. Google revealed that SynthID verification within the Gemini app has been used over 20 million times since its introduction in November. The company also confirmed plans to bring C2PA verification directly to the Gemini app in the near future.
