Overview
Nano Banana is Google DeepMind’s first widely available AI image generation model, built on Gemini 2.5 Flash (launched in August 2025). Nano Banana is optimized for precise prompt adherence and speed. It’s Google’s most efficient image model, built for high-volume tasks where turnaround time matters. While it’s limited to 1K resolution compared to its successors, the output quality is strong for everyday creative work — portraits, landscapes, product shots, and stylized scenes all land well. As a reasoning model, it understands natural language descriptions and the relationships between scene elements more deeply than standard diffusion models. You can describe complex scenes in plain language and get results that reflect your intent accurately, without needing to engineer prompts in a specific format.Getting Started
- Go to Image Generation — Navigate to krea.ai/image or select from the dropdown menu.
- Select Nano Banana — Open the model picker and choose Nano Banana from the Intelligent Models section.
- Write your prompt — Describe your subject clearly. Nano Banana handles a wide range of styles and subjects well.
- Add reference images (optional) — Upload images to guide style or composition.
- Choose your aspect ratio — Select portrait, landscape, or square.
- Generate — Click Generate. Nano Banana is fast and will return results quickly.
- Iterate — Adjust your prompt and regenerate, or bring your result into the Edit or Enhancer tool.
At a Glance
| Feature | Detail |
|---|---|
| Speed | Fast (3/3) |
| Credits | ~30 per generation |
| Underlying model | Gemini 2.5 Flash Image (Google DeepMind) |
| Resolution | Up to 1K |
| Reference image support | Yes |
| Best at | General use, precise prompt following, speed |
When to Use Nano Banana
| Use When | Avoid When |
|---|---|
| You need fast, reliable results for general tasks | You need output at 2K or 4K resolution |
| You’re iterating through many ideas quickly | You need to support more than a few reference images |
| Prompt adherence is important but budget is a consideration | You need advanced image editing capabilities |
| You want a solid, versatile model for most everyday tasks | You’re working on highly specialized or complex multi-element scenes |
Common Use Cases
- Portraits and people: Realistic human figures with accurate lighting and expression
- Landscapes and environments: Detailed natural and urban scenes
- Concept exploration: Fast iteration across many creative directions
- Stylized imagery: 3D figurines, artistic renders, mixed-style compositions
- General creative work: Anything that doesn’t require 4K output or heavy editing
Prompting Tips
| Tip | Example |
|---|---|
| Write naturally — no need for special syntax | Describe the scene the way you’d explain it to a person |
| Describe relationships between elements | ”A dog sitting to the left of a woman on a park bench” |
| Be specific about lighting and mood | ”Overcast afternoon light, muted palette, quiet atmosphere” |
| Specify artistic style if it matters | ”3D figurine style,” “editorial photography,” “flat illustration” |
| Use reference images to anchor style | Upload a visual reference rather than trying to describe a style in words |