Google Gemini 2.5 Flash Image brings a new 'bananas' upgrade to AI image generation
Google has unveiled Gemini 2.5 Flash Image, the viral ‘nano-banana’ AI model, bringing major upgrades to photo editing with better character consistency, multi-turn edits, and built-in watermarks.
- Aug 26, 2025,
- Updated Aug 27, 2025 2:16 AM IST
Google has officially rolled out Gemini 2.5 Flash Image, an upgraded AI image editing model that promises sharper, more consistent photo edits. The model had been teased for weeks under the quirky codename “nano-banana”, which went viral after surfacing anonymously on the crowdsourced evaluation platform LMArena. Testers speculated it was a Google project in disguise, and on Tuesday the company confirmed the rumours.
The update is now available in the Gemini app for both free and paid users, as well as through the Gemini API, Google AI Studio, and Vertex AI.
Built by Google DeepMind, Gemini 2.5 Flash Image is designed to fix one of the most frustrating problems in AI photo editing, maintaining character likeness and scene consistency. While rivals like ChatGPT and xAI’s Grok can sometimes distort faces or alter backgrounds when making simple edits, Google says its new system keeps people, pets, and objects looking true to life, even after multiple changes.
Examples highlighted by Google include reimagining a person with a 60s beehive haircut, dressing a dog in a tutu, or turning a butterfly into the design for a dress. Users can also blend photos, for instance, asking Gemini to merge a picture of a woman with one of a dog so they appear cuddling on a basketball court. Another feature lets you apply the artistic style of one image to an object in another, such as redesigning rain boots inspired by flowers.
Multi-turn editing is also supported, meaning users can refine an image step by step. Google demonstrated this with an empty room: the AI first repainted the walls, then added furniture like a bookshelf, couch, and rug while keeping the scene intact.
The company acknowledged that such powerful tools could be misused, so every AI-generated image will carry both a visible “AI” watermark and Google’s invisible SynthID marker to indicate its artificial origin.
According to LMArena rankings, Gemini 2.5 Flash Image is currently the highest-rated AI image editing model worldwide. Google is positioning the release as a key step in competing with OpenAI’s image tools, aiming to draw more users to its Gemini platform.
For Unparalleled coverage of India's Businesses and Economy – Subscribe to Business Today Magazine
Google has officially rolled out Gemini 2.5 Flash Image, an upgraded AI image editing model that promises sharper, more consistent photo edits. The model had been teased for weeks under the quirky codename “nano-banana”, which went viral after surfacing anonymously on the crowdsourced evaluation platform LMArena. Testers speculated it was a Google project in disguise, and on Tuesday the company confirmed the rumours.
The update is now available in the Gemini app for both free and paid users, as well as through the Gemini API, Google AI Studio, and Vertex AI.
Built by Google DeepMind, Gemini 2.5 Flash Image is designed to fix one of the most frustrating problems in AI photo editing, maintaining character likeness and scene consistency. While rivals like ChatGPT and xAI’s Grok can sometimes distort faces or alter backgrounds when making simple edits, Google says its new system keeps people, pets, and objects looking true to life, even after multiple changes.
Examples highlighted by Google include reimagining a person with a 60s beehive haircut, dressing a dog in a tutu, or turning a butterfly into the design for a dress. Users can also blend photos, for instance, asking Gemini to merge a picture of a woman with one of a dog so they appear cuddling on a basketball court. Another feature lets you apply the artistic style of one image to an object in another, such as redesigning rain boots inspired by flowers.
Multi-turn editing is also supported, meaning users can refine an image step by step. Google demonstrated this with an empty room: the AI first repainted the walls, then added furniture like a bookshelf, couch, and rug while keeping the scene intact.
The company acknowledged that such powerful tools could be misused, so every AI-generated image will carry both a visible “AI” watermark and Google’s invisible SynthID marker to indicate its artificial origin.
According to LMArena rankings, Gemini 2.5 Flash Image is currently the highest-rated AI image editing model worldwide. Google is positioning the release as a key step in competing with OpenAI’s image tools, aiming to draw more users to its Gemini platform.
For Unparalleled coverage of India's Businesses and Economy – Subscribe to Business Today Magazine