Google DeepMind has rolled out a new image‑editing model inside its Gemini app, called Nano Banana. The update promises to keep a person or pet’s likeness intact while swapping outfits, backgrounds, or styles. Users can drag a photo into Gemini, describe a change, and the model preserves the subject’s identity across edits.

Earlier this year, Gemini introduced native editing, but users complained about subtle mismatches. The latest iteration focuses on consistency, letting you blend multiple images or apply a texture from one picture to another without distorting the original subject. After each edit, the app stamps a visible watermark and an invisible SynthID tag to flag AI origin.

Practical use cases span from personal selfies—trying a 60s beehive or a tutu on a dog—to commercial needs like background replacement or style transfer. The model also supports multi‑turn editing, letting users paint a room, then add furniture, all while the unchanged parts stay true to the source.

All images edited in Gemini carry a visible watermark and a SynthID digital watermark, ensuring creators can prove AI involvement. With Nano Banana, developers and hobbyists gain a versatile tool that blends artistic freedom with reliable subject fidelity, making it a solid addition to the suite of AI‑powered media tools.