Google DeepMind Supercharges Gemini’s AI Image Editing with “Nano Banana” Model

Date: August 26, 2025

Google DeepMind today unveiled a groundbreaking enhancement to its Gemini app’s image editing capabilities: the integration of the “Gemini 2.5 Flash Image” model—playfully dubbed “Nano Banana.” This update marks a major leap forward, offering users sharper control over edits while preserving the recognizable essence of people and pets in their photos.

What’s new:

Character Consistency: Previous AI tools often altered subtle features, making edited subjects appear off. The new model keeps appearances intact—even with dramatic edits like changing outfits, backgrounds, or eras.
Multi-image Fusion: Users can now blend two or more photos into a single cohesive scene—such as combining a photo of you and your pet into one seamless image.
Multi-turn & Prompt-Based Editing: Perform incremental edits—like painting a room, then adding furniture—with natural language instructions. You can also apply a style from one image (e.g., flower petals) to an object in another (e.g., designing a dress).
Developer Access & Pricing: The model is available now via the Gemini app, Gemini API, Google AI Studio, and Vertex AI. Developers and enterprises can use it directly for creative applications. Pricing starts at $30 per million output tokens—equating to approximately $0.039 per image.
Transparency with AI Artifacts: All edited images feature a visible watermark plus an invisible SynthID tag to clearly indicate they were AI-generated.

Why it matters:

The update addresses a common pain point in AI image editing—preserving subject fidelity. By maintaining character likeness through natural-language edits and image blending, Gemini now supports both creative freedom and authentic results. It also cements Google’s position in the AI editing space, responding to growing user demand for intuitive, reliable image transformation tools.