Nano Banana revolution at Google: The artificial intelligence that shook the internet is officially live
It was revealed that the new artificial intelligence model, which recently appeared on the internet under the name "Nano Banana" and reached the top of the artificial intelligence benchmarking platform LMArena, was developed by Google DeepMind.
Google announced that it has integrated the model into its Gemini app.
Users who have tried Nano Banana claim it could revolutionize image editing. Some believe these AI models could replace photo editing apps like Photoshop.
CONSISTENT AND REALISTIC REGULATIONS
Google added basic image editing capabilities to Gemini earlier this year. However, with previous systems, it was common for the image to change unexpectedly with each edit. It took a lot of experimentation and a series of commands to get the image edited to the user's liking.
The new model, officially known as Gemini 2.5 Flash Image, largely addresses this issue. According to Google, the model provides unprecedented consistency in edits and can remember details from the original image during modifications.
This feature allows users to upload a photo of a person and, for example, change their outfit or reimagine them in different styles. Even with multiple edits, the results still look like the person in the original photo.
COLLAGE ABILITY
The new model isn't limited to editing a single image. Users can combine multiple photos to create new composite images. In Google 's example, a separately uploaded photo of a woman and a dog was transformed into a new scene featuring the dog being held in a hug.
In addition, the model can produce original images according to users' instructions by making more abstract combinations.Google announced that the new editing tool is rolling out to all Gemini users as of Tuesday. The model will also be available to developers soon through the Gemini API, AI Studio, and Vertex AI.
Nicole Brichtova, a product lead for Google DeepMind, said the model is designed specifically for consumer use cases, including visualizing home and garden projects and combining multiple references (e.g., a sofa, a living room photo, and a color palette) into a single design. "This update makes editing much more seamless, and the outputs can be used for anything you want," Brichtova said.HOW TO USE?
Gemini 2.5 Flash Image is available for a fee. Developers receive token-based payments via the API ($0.039 per image), while individual users pay through subscription plans like Google AI Pro, which provide access to this model. Subscribers using the Google AI Pro plan pay around $20 per month. This plan offers access to some advanced models and features. However, it's important to note that there's no separate pricing for individual users for Flash Image within the Gemini app. Access is provided through the plan included with the model. Those who want to try the new model can download the "Gemini" app from the Google Play Store or App Store. Then, you can subscribe by selecting the Gemini Advanced / AI Pro upgrade from the menu or settings. The web version is available at "gemini.google." You can start a subscription by selecting "Upgrade to Gemini Advanced" in the bottom left corner or in your profile.The app's workflow is similar to other AI tools. After selecting the "upload image" option, you're prompted for editing. For example, you can issue commands like "Change this person's outfit to a '90s style" or "Remove the background and replace it with a beach."
Competition is heating up
Google's move follows OpenAI's announcement of visual generation capabilities with GPT-4o in March. At the time, ChatGPT saw a surge in usage, and its "Studio Ghibli"-style images generated a significant buzz on social media. Meta also announced last week that it would acquire visual models from Midjourney. Germany-based Black Forest Labs' FLUX models are still among the highest-performing solutions in the industry. Google believes Gemini's advanced visual generation capabilities could help the company close the user gap with OpenAI. ChatGPT currently reaches 700 million weekly users, while Google CEO Sundar Pichai announced Gemini's 450 million monthly users during a July earnings call.PRECAUTIONS AGAINST DEEPFAKE RISK
Like Google's other AI image models, every image generated with Gemini 2.5 Flash Image features the word "AI" in the corner. The images also feature an invisible SynthID digital watermark, which remains visible against the background even when edited. This is Google's way of protecting against the threat of deepfakes. However, the company acknowledges that such markings can be overlooked in images that are quickly shared on social media and used for disinformation. Problems with Gemini's image generation have been raised in the past. The company was forced to apologize and withdraw the system, particularly for historically inaccurate images of people (for example, depicting Elon Musk as Black). The company now states that it has developed a more balanced approach.Google's terms of use explicitly prohibit the creation of unauthorized intimate content. This is where Google differentiates itself from its competitors. For example, Grok, developed by Elon Musk's xAI company, allowed the creation of sexually explicit artificial images resembling celebrities.
ntv
