Several months of testimony saw Google CEO Sundar Pichai and other top executives take the stand
Google's new AI photo-editing tool, the Gemini 2.5 Flash Image, was playfully codenamed 'Nano Banana'. The model's key ability is maintaining a subject's look across various edits, a feature inspired by CEO Sundar Pichai’s dog, Jeffree. AFP News

In an unexpected twist for the world of artificial intelligence, Google's latest breakthrough was unveiled not with a flashy presentation but through a playful social media post from CEO Sundar Pichai.

The company's new AI photo-editing tool, known by the quirky codename 'Nano Banana', was showcased with a series of imaginative images featuring Pichai's own dog. This demonstration has the tech world buzzing.

An AI Model with a Flashy New Name

Google CEO Sundar Pichai recently took to X (formerly Twitter) with a cryptic post of three banana emojis, sparking immediate speculation. The post was a nod to the now-viral codename 'Nano Banana', which was revealed to be a new AI tool.

A few hours later, Google officially introduced the AI as a new photo-editing tool, which lets users edit and modify images with a simple prompt. Although internally codenamed 'Nano Banana', the new tool has been publicly launched as Gemini 2.5 Flash Image.

Now integrated into the Gemini app, this generative AI model promises high-quality, context-aware images. Its key feature is the ability to maintain the original subjects' look across multiple generations.

A Canine Catalyst for Google's Next Big Thing

To demonstrate the new AI image editor's capabilities, Google CEO Sundar Pichai used X (formerly Twitter) to share a series of lighthearted pictures of his dog, Jeffree. The images were generated with the new tool to mark International Dog Day.

Pichai used the new AI to reimagine Jeffree, showing him surfing waves, donning a cowboy hat, turning into a superhero, and even dressing as a chef. Throughout the different themes, the new AI model was able to keep the dog's original look recognisably consistent.

In a post on X, Pichai announced that the new image editing model was now 'rolling out in @Geminiapp - and yes, it's (Banana emojis). Top of @lmarena's image edit leaderboard, it's especially good at maintaining likeness across different contexts. Check out a few of my dog Jeffree in honor of International Dog Day - though don't let these fool you, he definitely prefers the couch:).'

A Breakdown of the Gemini 2.5 Flash Image Model

This new image editor is an evolution of the image generation features Google introduced in the Gemini app earlier this year. According to the company, a frequent complaint with AI-generated photos is that an edited picture of a person or pet doesn't quite look like the original. This is a problem the 'Nano Banana' update was specifically created to solve.

To use the new model, simply upload a picture and provide a prompt. Gemini's AI can then apply a wide range of transformations while maintaining the subject's recognisable appearance.

Going beyond simple, prompt-based edits, the new AI image-editing model now includes two other notable capabilities. The first is multi-turn editing, which allows users to build scenes in stages. The second is design mixing, which allows you to borrow patterns and textures from one photo and apply them to another.

For example, you could use the wings of a butterfly as a fabric print for a dress or give a pair of rain boots the texture of flower petals. Additionally, the new image generation model can blend multiple photos into a single, seamless picture. Users can even feed their edited images back into Gemini to generate animated videos.

A Global Rollout with a Focus on Transparency

The Gemini 2.5 Flash Image model was rolled out to users around the world on 26 August, making it accessible within the Gemini app for both free and paid accounts. To ensure transparency, every image created by the AI will feature a clear, visible watermark in addition to Google's invisible SynthID digital watermark. This makes it obvious that the pictures were generated by AI.

For developers and enterprises, Google is also making the new Gemini 2.5 Flash Image model available through the Gemini API, Google AI Studio and Vertex AI. The pricing is set at $30 (£22.25) per million output tokens, meaning each generated image costs around $0.039 (£0.029).