Google has officially released the much-rumored "Nano-Banana" AI image generator for the Gemini app. Officially called "Gemini 2.5 Flash Image," the tool is available in the app for both free and paid subscribers. It's also accessible through Google AI Studio, the Gemini API, and Vertex AI.
This model, developed by Google DeepMind, attracted attention by achieving top scores on the LMArena benchmark. Even before its release, NanoBanana garnered attention for its impressive editing capabilities, garnering attention through demonstrations in online communities and social media.
The core of the update is "maintaining character consistency." Google explained that it's designed to ensure that people in photos are recognized as the original person even when their hairstyle or clothing changes. For example, even when dressing a pet in a costume, compositing multiple photos, or applying a specific style to another object, the person or object in the image remains consistent and unaltered. Multi-step editing, such as changing the interior of a room, is also supported, enhancing usability.
The new model also produces results that look like real photos. The Nano Banana sample has already captured public attention for its realistic image quality. Free users can edit up to 100 images per day, while paid users can edit up to ten times as many images.
Google has also strengthened its security measures. All created and edited images now feature a prominent "AI" watermark and Google's SynthID digital watermark. This is to prevent potential abuses like deepfakes and the spread of misinformation.
The AI image generation competition is rapidly intensifying. While OpenAI took the lead by unveiling a powerful generator earlier this year, Google is now on the offensive with Gemini. The release of Nano Banana is considered a significant attempt to push the boundaries of AI, focusing on image editing and compositing capabilities.
