Model + Tool
A Nano Banana image-to-image editor uses Google's Gemini 2.5 Flash Image model to edit existing images via natural-language instructions, with multi-image composition and conversational chain edits that keep state across multiple turns.
Type "add a hat" or "make the sky stormy" — Nano Banana edits the image without you re-prompting from scratch.
Each edit builds on the previous one — "now add sunglasses, then change to evening lighting" stays coherent.
Upload a person + an outfit + a setting; Banana composes them into one image with consistent lighting.
Subject stays recognizable across edits without explicit face-lock prompting.
Yes — that's exactly what Nano Banana does. "Make the sky stormy" or "add sunglasses" both work.
Usually no — it figures out where to apply the change. For very specific regions, the Inpaint tool with masking is more precise.