Google upgrades Gemini image generation to Nano Banana 2 with 4K support

4K rendering, multimodal workflows, and C2PA-compatible SynthID watermarking built in

Serving tech enthusiasts for over 25 years.
TechSpot means tech analysis and advice you can trust.

TL;DR: Google is replacing its existing image generation engine across the Gemini ecosystem with a new model, Nano Banana 2, officially known as Gemini 3.1 Flash Image. The update delivers substantial gains in realism, speed, and scene complexity, making it the default generator for all Gemini image modes, including Fast, Thinking, and Pro.

The move represents a step toward unifying Google's image AI lineup under a single, high-performance model. Since the debut of the first Nano Banana in August 2025, Gemini users have created millions of images globally while steadily pushing the models' creative limits using Gemini apps and tools, such as the AI-powered video editor Flow.

Nano Banana 2 draws on architectural improvements from last year's Nano Banana Pro, but with faster output and more nuanced control.

Images can now be generated at resolutions from 512 pixels up to 4K and in multiple aspect ratios, making the tool versatile for both casual use and professional media production. Google says the new model supports character consistency across as many as five subjects and consistent rendering of up to fourteen objects, enabling more coherent storytelling in complex prompts.

The company also emphasizes advances in lighting, texture, and fine detail. Generated visuals now handle reflective surfaces and dynamic scenes with greater precision. These visual refinements, combined with higher throughput, effectively close part of the gap between consumer-facing image models and professional-grade rendering systems.

As part of the update, Nano Banana 2 becomes deeply embedded throughout Google's platform stack. It will replace older models within the Gemini app and extend to external products such as the Flow video tool, Google Search (via Lens integration), and AI Mode across 141 countries on both mobile and desktop.

For Google AI Pro and Ultra users, Nano Banana Pro is still available for detailed image tasks via manual regeneration.

For developers, Nano Banana 2 is now offered in preview through the Gemini API, Gemini CLI, the Vertex API, and AI Studio. It also integrates with Google's developer environment Antigravity, first introduced with Gemini 3 in late 2025, allowing direct experimentation with image-to-code and multimodal workflows.

All images generated through Nano Banana 2 carry a SynthID watermark, Google's invisible signature designed to identify AI-created content. These watermarks are compatible with the industry-wide C2PA standard, which also includes contributors like Adobe, Microsoft, OpenAI, and Meta.

// Related Stories

• The Interim Computer Museum puts 28 classic computer systems on the internet for you to try
• We found 8 Nano Banana Pro images that show Google is the new benchmark in AI art

Google reports that SynthID verification has already been used more than 20 million times since its rollout in the Gemini app last November – a sign that content-provenance features are becoming a routine expectation in creative AI systems.