Google has launched a significant upgrade to its AI image generation capabilities with the release of Gemini 3 Pro Image, widely known as “Nano Banana Pro.” This new model, built on Gemini 3, aims to resolve key weaknesses in AI-generated images—particularly the difficulty in creating legible text—while also introducing higher resolution outputs and advanced image manipulation features.
Enhanced Capabilities and Accessibility
The update focuses on improvements in three main areas: text generation, reference image handling, and resolution. Users can now generate images with clearer, more accurate text, a feature that has been problematic for previous AI models. Gemini 3 Pro Image can utilize up to 14 reference images, allowing for more detailed and complex compositions. Finally, the model supports resolutions up to 4K, delivering higher-quality visuals.
The update is currently available globally within the Gemini app. To access the new model, select “Create images” (indicated by a 🍌 emoji) and switch to the “Thinking model” option within the prompt window.
Cost and Access Options
While Nano Banana Pro is accessible for free, users will encounter generation limits more quickly than those with paid subscriptions. Google AI subscriptions start at $20 per month. Once free credits are exhausted, the system reverts to the standard model. The model is also integrated into the Gemini API, Google AI Studio, and Flow.
Addressing AI Image Artifacts
AI image generators have historically struggled to render text accurately, resulting in distorted or illegible characters. OpenAI’s GPT-Image-1 made early progress in this area, but reliability remained an issue. Google claims Nano Banana Pro significantly improves text rendering, potentially enabling the creation of AI-generated infographics with readable text. The integration of Gemini with Google Search (and AI Mode for Pro subscribers) further enhances this capability.
Watermarks and AI Detection
Google is also altering its approach to watermarks. High-tier subscribers (Ultra plan at $250/month) can now remove AI-generated watermarks, making it harder to distinguish between AI-created and real images. Despite this, Google asserts that its SynthID system—an invisible watermark embedded in generated content—will allow Gemini to identify AI-created images if uploaded for verification. The effectiveness of SynthID remains to be seen.
The “Nano Banana” Legacy
The name “Nano Banana Pro” originated from the previous iteration, Gemini 2.5 Flash Image, released earlier this summer. Early testers praised its character consistency, meaning the model could edit existing photos without distorting the people or objects within them. This playful naming convention, reinforced by Google executives, quickly caught on and became the model’s unofficial moniker.
Gemini 3: A Broader Upgrade
The launch of Nano Banana Pro is part of a larger Gemini 3 update. These changes aim to deliver a more intelligent and less biased AI experience for Gemini users. This includes improvements across multiple aspects of the AI system, not just image generation.
In conclusion, Google’s Nano Banana Pro represents a step forward in AI image generation, addressing long-standing issues with text clarity, resolution, and watermark transparency. The update underscores the rapid evolution of AI tools and raises questions about the increasing difficulty in distinguishing between AI-created and real content.
