Google’s “Nano Banana” Lands in Gemini: What the New Image Model Can Do

Google’s “Nano Banana” Lands in Gemini: What the New Image Model Can Do

Introduction

Google has quietly rolled out a sophisticated image‑editing model nicknamed “nano banana” – formally Gemini 2.5 Flash Image – inside the Gemini app. The model first surfaced anonymously in AI benchmarking platforms under the pseudonym “nano banana” and quickly topped user rankings for its striking consistency and versatility. Now Google confirms that it is the native image capability of Gemini 2.5 Flash and is available to everyone through the Gemini web and mobile apps as well as to developers via Google AI Studio, the Gemini API and Vertex AI.

Why “Nano Banana” Matters.

Traditional image‑editing models struggle to keep faces and objects stable across multiple changes; altering a shirt color often distorts the subject’s features. Nano banana aims to solve this with improvements that:

- **Maintain character consistency** so a person, pet or product looks the same across edits and scenes.
- **Blend multiple images** seamlessly to create new compositions, such as combining photos of you and your pet without mismatched lighting or proportions.
- **Follow complex, multi‑turn instructions** so you can make a series of edits without the image “drifting” off topic.
- **Understand world knowledge** to interpret natural‑language prompts more accurately.

These advances stem from Gemini’s broader architecture and build on the feedback Google received after launching the Gemini 2.0 Flash generator earlier in 2025. Users wanted higher‑quality images and better creative control, which this update addresses.

Key Features and Examples.

**Costume and background changes:** You can upload a selfie and ask Gemini to place you in different outfits, professions or eras, and the model preserves your facial details and proportions. Similarly, changing the background of a room or putting yourself in a new location keeps the subject intact.

**Photo blending:** Nano banana fuses multiple images; for instance, you can merge a picture of yourself with your dog to create a joint portrait with realistic lighting.

**Step‑by‑step editing:** You can redesign a room by painting walls, adding furniture and decorations one step at a time, with each change respecting previous edits.

**Design mixing:** Borrow colors and textures from one image and apply them to objects in another; for example, use flower petal patterns to design clothing.

Google says all images created or edited in the Gemini app include a visible watermark and an invisible SynthID digital watermark to indicate AI‑generated content.

Availability and Pricing

The nano banana model is now live for both free and paying users in the Gemini app worldwide. Developers can access it via the Gemini API, Google AI Studio and Vertex AI; pricing is $30 per one million output tokens, and each image costs about 1290 tokens, or roughly $0.039. The model remains in preview for now but will stabilize in the coming weeks.

The Banana Hype and Meme.

Weeks before the official announcement, LMArena and other testing platforms were buzzing about an anonymous image‑editing model that produced remarkably consistent results. Google DeepMind dropped hints with banana‑themed memes, and CEO Sundar Pichai posted three banana emojis on X, teasing the tool and fueling speculation. When the company finally acknowledged the model, it confirmed that “nano banana” is simply the colloquial name for Gemini 2.5 Flash Image.

Outlook

Nano banana underscores how quickly generative image tools are maturing. By addressing the biggest pain points — identity consistency and multi‑turn editing — and making the technology accessible through consumer apps and APIs, Google is positioning Gemini to compete directly with OpenAI’s image models. As the model moves from preview to general release, expect more integrations in Google’s products and a growing ecosystem of third‑party apps built on top of this creative AI engine.