Nano Banana 2 AI Image Generator
Nano Banana 2 is Google's Gemini 3.1 Flash Image model: Pro-level quality at Flash speed, up to 14 reference images, legible text and 4K output.
Generated image will appear here
Enter a prompt and click Generate
Prompts to try with Nano Banana 2
See Nano Banana 2 examples built around multi-reference editing, character consistency, style transfer and legible in-image text. Click an example to load its prompt into the generator.
What is Nano Banana 2?
Nano Banana 2 is Google's Gemini 3.1 Flash Image model — the successor to the original Nano Banana, released in February 2026. It brings the world knowledge, reasoning and visual quality of Nano Banana Pro down to Flash speed: fast iteration, precise instruction following, real-time Google Search grounding, legible text rendering and consistent subjects. On GenPix you can run Nano Banana 2 online in the prompt box above — text-to-image or reference-based editing, up to 4K, from 5 credits per image.
Gemini 3.1 Flash Image
Google's newest image model: Pro-level quality at Flash speed.
512px to 4K output
Square, portrait, landscape and ultra-wide — sharp at any ratio.
SynthID provenance
Every generated image carries Google's invisible watermark.
What Nano Banana 2 is good at
The tasks Nano Banana 2 handles best — and when to reach for it instead of a slower, heavier model.
Multi-reference blending
Feed it several photos — person, product, background — and every subject stays recognizably itself in the final image.
Character consistency
Keep up to 5 characters and 14 objects faithful across a whole workflow, so storyboards and campaigns stay on-model.
Precision text rendering
Headlines, labels and packaging copy come back legible and correctly spelled — and can be translated in place.
Real-time knowledge
Grounded in Gemini's knowledge base and live web search, so infographics, charts and diagrams reflect current facts.
Sketch and draft rendering
Turn a rough pencil sketch or wireframe into a finished, photoreal render in a single pass.
Fast iteration
Flash-speed edits come back in seconds, so you can explore many directions in one sitting.
Why use Nano Banana 2 for reference-driven images?
When the result has to keep several things true at once — your product, your model, your room, your style — Nano Banana 2 is built for it. Multiple references go in, one consistent image comes out. Because it shares the same prompt box as Nano Banana Pro, GPT Image 2.5 and Seedream 5.0 Pro, you can also compare engines on the identical prompt with one click.
Pro quality, Flash speed
Nano Banana Pro-level output without the wait.
Same prompt, other models
Switch to Pro or GPT Image 2.5 in one click.
Up to 4K
Sharp enough for print, packaging and posters.
How to use Nano Banana 2
Treat the prompt like a compact creative brief. State the exact content, explain where it belongs, and use reference photos whenever identity or product details must stay consistent.
- 1
Describe the shot or upload references
Type what you want, or switch to Image to image and add up to 14 reference photos. Pick an aspect ratio and a resolution — 1K for drafts, 4K for final output.
- 2
Give every reference a role
Number your inputs by job: photo 1 is the subject, photo 2 is the outfit, photo 3 is the background. Say which is which in the prompt.
- 3
State what must stay the same
List the things to preserve — face, logo, label, layout — before describing the change. Constraints keep the edit anchored.
- 4
Steer style with one strong reference
Pick a single style image instead of describing style in words. The model transfers palette and brushwork more reliably from an example.
- 5
Iterate in the same conversation
Ask for follow-up edits — a new background, new lighting, translated text — and the subject stays consistent from image to image.
Explore more AI image models
One prompt box, several top models — switch whenever a shot calls for a different engine.
Nano Banana 2 FAQ
What people ask before generating their first image with Nano Banana 2.
Blend your references with Nano Banana 2
Upload your photos, describe the result, and let the references do the heavy lifting — characters, products and places that stay themselves, in up to 4K.