Google reveals “nano-banana” is its new Gemini 2.5 Flash Image model

Google has confirmed that the mysterious "nano-banana" model is actually its new Gemini 2.5 Flash Image system, designed for image generation and editing through artificial intelligence. The system allows users to create and modify visual content using natural language instructions, expanding the capabilities of the previously introduced Gemini 2.0 Flash model.

Among the main features of the new model is the ability to maintain visual consistency of characters or objects across multiple images. This function proves useful for developers who need to create coherent content, such as product promotional materials or visual narratives requiring recognizable characters in different scenarios.

The system also incorporates localized editing functions that allow specific modifications to particular parts of an image. Users can remove unwanted elements, change subject poses, apply selective blur effects, or add color to black and white photographs through simple text commands.

A notable feature is the ability to merge multiple input images into a single composition. This function allows combining objects from different photographs, applying specific color schemes to interior spaces, or creating photorealistic scenes that integrate elements from various visual sources.

The model uses Gemini's general knowledge to interpret real-world contexts, enabling more semantically accurate image generation. This integration facilitates the creation of interactive educational content and understanding of hand-drawn diagrams.

Gemini 2.5 Flash Image is available through the Gemini API, Google AI Studio, and Vertex AI for enterprises. The established price is $30 per million output tokens, with each image equivalent to 1,290 tokens, representing a cost of $0.039 per generated image.

Google has developed several demonstration applications showcasing the model's capabilities, including photo editors, interior design tools, and collaborative drawing systems. These applications are available as customizable templates in Google AI Studio.

All images created or edited with this model include an invisible SynthID digital watermark, allowing identification of AI-generated or modified content. The company has established partnerships with platforms like OpenRouter.ai and fal.ai to expand developer access to the new model.

Key Points:

• Google confirms "nano-banana" is its new Gemini 2.5 Flash Image model for AI image generation and editing
• The system allows image editing through natural text commands, including element removal and pose changes
• Maintains visual consistency of characters and objects across multiple generated images
• Can merge multiple input images into a single photorealistic composition
• Available through Gemini API, AI Studio, and Vertex AI
• Includes invisible SynthID digital watermark to identify AI-generated content