Google introduced Gemini 2.5 Flash Image, a new model dubbed "nano-banana" that brings a range of advanced features to image generation and editing, including seamlessly blending multiple images, maintaining character consistency, and utilizing natural language cues for precise editing, providing powerful tools for creative workers and enterprise applications.
Google has taken a major step forward in image generation technology with the launch of Gemini 2.5 Flash Image, a new model nicknamed the "nano-banana" that brings a host of advanced features that enable users to seamlessly blend multiple images, maintain character consistency across scenes, and perform precise image editing with natural language cues. and accurate image editing through natural language cues. In addition, the model integrates Gemini's extensive world knowledge to support image generation and editing capabilities.
Since the release of Gemini 2.0 Flash earlier this year, user feedback has driven a number of improvements, particularly the need for high-quality images and more creative control. Today, Gemini 2.5 Flash Image is available through the Gemini API, Google AI Studio (for developers), and Vertex AI (for enterprise applications), priced at $30 per million output tokens and approximately $0.039 per image.
To streamline the creation process, Google AI Studio has undergone a major update, particularly in the area of "Build Mode". Users can customize the AI-powered app's ability to explore models, remix existing templates, or generate new ideas with simple prompts. Completed projects can be easily deployed or saved on GitHub.
A key advancement in the model is the ability to maintain character consistency across images. This means that the same character can be set up in different environments, or show a product from multiple angles, while keeping its unique characteristics intact. In addition, Gemini 2.5 Flash Image's ability to closely follow visual templates is attractive to developers working on projects such as real estate listing cards, employee badges, or full-scale product mockups, all of which can be generated from a single design template.
The model also features cue-based image editing, allowing users to make precise and targeted transformations. For example, users can blur backgrounds, remove unwanted elements or adjust poses with simple prompts, making complex edits easy to follow.
In addition, the model's integrated knowledge of the world allows it to function beyond aesthetics. It is able to support applications such as interactive educational tools that can interpret hand-drawn diagrams and respond to real-world problems while performing complex editing commands in a single step.
Another notable feature is multi-image fusion, which allows users to blend various input images together to produce a coherent and realistic output. This is particularly useful in product design and marketing, where visual storytelling plays a key role.
Using Gemini 2.5 Flash Image is made easy with comprehensive developer documentation and previews available through the Gemini API and Google AI Studio. The model has partnered with OpenRouter.ai to enable it to reach over three million developers, making it the first image generation model on the platform. Additionally, the partnership with fal.ai will make the technology available to the wider development community.
In order to maintain transparency and accountability, all images generated or edited via Gemini 2.5 Flash Image will include an invisible SynthID digital watermark labeled as AI-generated or edited.
Improvements continue to be a work in progress, with a focus on enhancing long text rendering, character consistency and ensuring accuracy of image details. User feedback is critical, and developers can share insights on forums or social media platforms. As anticipation for these new features continues to grow, the potential for innovation in the image generation space seems endless.