Model + Tool
A Nano Banana text-to-image generator uses Google's Gemini 2.5 Flash Image model (codename: Nano Banana) to create images from natural-language prompts, with industry-leading instruction-following on long, complex prompts and multi-image composition.
Nano Banana handles long, multi-clause prompts ("a corgi wearing a chef hat, holding a baguette, in a Parisian cafe at sunset") better than any peer.
Banana 2 for fast iteration, Pro for hero output with strongest prompt adherence.
Upload reference images (person + outfit + setting); Nano Banana composes them into one coherent generation.
"Now add a hat", "now change the lighting" — Nano Banana keeps the conversation state.
It was the internal codename Google used during testing before launch — it stuck.
Nano Banana is specifically tuned for natural-language editing and multi-image composition. Imagen focuses on raw text-to-image quality.