To use Nano Banana Pro, open the Gemini app, select Create images, choose the Thinking model, and pick Nano Banana Pro. Write a detailed prompt and optionally upload up to 14 reference images for consistent characters or style. Generation takes 8-15 seconds; free-tier users get a limited daily allowance before falling back to the standard model.
A step-by-step guide to generating your first image with Nano Banana Pro -- from opening Gemini to writing prompts that use its 4K resolution and text-rendering strengths.
How to Use Nano Banana Pro: Step-by-Step Guide
Nano Banana Pro is the nickname that stuck for Gemini 3 Pro Image, Google DeepMind's image-generation model built into the Gemini app. This guide walks through actually generating an image with it -- from opening the app to writing prompts that take advantage of its native 4K output and accurate in-image text. For pricing details, see our Nano Banana Pro pricing breakdown; for a full verdict against Midjourney and DALL-E, read our Nano Banana Pro review.
Step-by-Step: Generating Your First Image
- Open Gemini. Go to gemini.google.com or open the Gemini app on desktop or mobile.
- Start image creation. Open the tools menu and select "Create images."
- Choose the Pro model. Select Gemini Thinking, then choose Nano Banana Pro under image generation -- this is what unlocks native 4K output and multi-image fusion instead of the standard, faster Nano Banana model.
- Write a detailed prompt. Describe the subject, action, setting, camera angle, lighting, and style in one clear sentence rather than a list of disconnected keywords.
- Add reference images (optional). Upload up to 14 images to guide character consistency, composition, or a specific visual style across a batch of generations.
- Generate. Thinking mode takes roughly 8-15 seconds per image since it reasons through the prompt before rendering, longer than the standard model but noticeably more accurate on complex scenes.
Writing Prompts That Actually Work
Nano Banana Pro responds best to prompts structured in layers rather than a flat list of adjectives: [subject + descriptive detail] doing [action] in [location/context], described with [composition or camera angle], [lighting/atmosphere], and [style or medium]. For example: "A weathered fisherman mending nets on a wooden dock at dawn, low-angle shot, warm golden-hour light, shot on 35mm film." Because the model has built-in reasoning, it holds up on genuinely complex requests -- multiple named subjects, specific in-image text, or layouts with several distinct elements -- better than models that generate directly from the raw prompt without that reasoning step.
Getting Accurate Text Inside Images
In-image text is one of Nano Banana Pro's clearest strengths over older models. To get it right, spell out the exact text in quotes within your prompt (e.g., a sign reading "OPEN 24 HOURS") rather than describing the sign generically, and keep the requested text short -- accuracy drops on long paragraphs the same way it does for most image models, just less severely.
Using Reference Images for Consistent Characters
The Few-Shot Design capability accepts up to 14 reference images in a single request, which is enough to lock a character's face, outfit, or a product's exact appearance across an entire batch of scenes. This matters most for anything produced as a series -- a comic page, a product line shot in multiple settings, or marketing assets that need to look like the same subject throughout.
Free Tier Limits and Access
The free Gemini tier includes a small daily allowance of Nano Banana Pro generations -- once you hit it, the app quietly falls back to the standard (non-Pro) Nano Banana model for the rest of the day. Google AI Pro ($19.99/month) and AI Ultra ($99.99/month) subscribers get substantially higher daily limits. Developers can also access the model directly through the Gemini API as gemini-3-pro-image-preview, billed per image rather than through a subscription -- see the full pricing table for exact per-image API costs.
Frequently Asked Questions
How do I access Nano Banana Pro?
Open the Gemini app or gemini.google.com, select "Create images" from the tools menu, then choose Gemini Thinking and pick Nano Banana Pro under image generation.
How many reference images can Nano Banana Pro use?
Up to 14 in a single request through its Few-Shot Design capability, enough to lock a character, product, or style consistently across a whole batch of generated images.
Why does Nano Banana Pro take longer to generate an image?
Thinking mode reasons through your prompt before rendering, which takes roughly 8-15 seconds -- longer than the standard Nano Banana model, but noticeably more accurate on complex scenes and in-image text.
What happens when I hit the free daily limit?
The Gemini app quietly switches you to the standard (non-Pro) Nano Banana model for the rest of the day. Google AI Pro and Ultra subscribers get substantially higher daily limits before that happens.
How do I get accurate text inside a generated image?
Put the exact text in quotes within your prompt (e.g., a sign reading "OPEN 24 HOURS") rather than describing it generically, and keep requested text short -- accuracy drops on long passages.
⭐ User Reviews
Be the first to review!