How to Use Nano Banana Pro: The Beginner's Guide (First Image, Reference Images, and the Free Quota)
CodexFuse Team · 2026-10-28T06:57:42.394Z · 16 min read
Nano Banana Pro is Google's Gemini 3 Pro Image model. Open the Gemini app, pick the Thinking model, write a detailed prompt with up to 14 reference images, and generate — free users get a small daily quota.
How Do You Use Nano Banana Pro?
To use Nano Banana Pro, open the Gemini app, select the Thinking model, choose Create images, type a detailed prompt, and tap Generate. You can attach up to 14 reference images to keep people, products, and style consistent, then pick a 1K, 2K, or 4K output. Nano Banana Pro is Google's Gemini 3 Pro Image model, so it is slower and more deliberate than the original Nano Banana — but it is the one to use when the image has to hold together: readable text, real faces, brand assets, or a factual infographic.
- Open the Gemini app on the web or your phone and sign in with your Google account.
- Select the Thinking model from the model dropdown above the prompt box.
- Choose Create images to start an image generation.
- Write your prompt starting with "create an image..." and describe the subject, lighting, and style in one clear sentence.
- Attach reference images if you need them (up to 14), choose your resolution, and tap Generate.
That is the whole loop. This Nano Banana Pro tutorial walks through each step in detail below, with the free quota, the reference-image system, and the paid line explained so you know exactly what you get before you pay for anything.
What Is Nano Banana Pro?
Nano Banana Pro is Google DeepMind's Gemini 3 Pro Image model, launched on November 20, 2025. It is built on Gemini 3 Pro, which means it reasons about your prompt before it draws — it plans the layout, checks real-world facts through Google Search, and then renders the image. Google describes it as its most advanced image generation and editing model, aimed at studio-quality results rather than quick fun edits.
The original Nano Banana is the Gemini 2.5 Flash Image model: fast, playful, and good for quick edits and experiments. Nano Banana Pro is the upgrade for work that has to be right. The table below shows the difference in one place.
| Nano Banana (original) | Nano Banana Pro | |
|---|---|---|
| Model behind it | Gemini 2.5 Flash Image | Gemini 3 Pro Image |
| Best for | Fast drafts, fun edits, quick ideas | Final assets, text-heavy images, brand work |
| Reference images | Yes, fewer inputs | Up to 14 images blended in one prompt |
| People consistency | Basic | Up to 5 people kept consistent |
| Resolution | Standard | 1K, 2K, and 4K output |
| Text in images | Good for short text | Legible text in multiple languages, including long passages |
| Speed | Fast | Slower — it thinks before it renders |
| Factual grounding | No | Connects to Google Search for real-world facts |
One naming warning before you go further: some third-party tools label this model "Nano Banana 2." Google's own name is Nano Banana Pro, also called Gemini 3 Pro Image, and that is the name to trust. Recent third-party coverage lists Nano Banana 2 and Nano Banana Pro as separate models, so treat any tool that uses the names interchangeably with care and check which model ID it actually calls.
Where Can You Use Nano Banana Pro?
Nano Banana Pro is available in several Google products, and the right entry point depends on what you are doing. Most beginners should start in the Gemini app. Developers should start in Google AI Studio.
| Where | Who it is for | What you get |
|---|---|---|
| Gemini app | Everyone | Create images with the Thinking model; free users get a limited quota, then fall back to the original Nano Banana |
| Google AI Studio | Developers | API access with the model ID gemini-3-pro-image; no visible watermark on outputs |
| Google Antigravity | Developers and designers | Rich UX layouts and mockups built with the Pro model |
| Google Flow | Filmmakers and creators | Pro image generation inside the AI filmmaking tool (Ultra subscribers) |
| Google Slides and Vids | Workspace customers | Pro images inside presentations and videos |
| NotebookLM | Subscribers | Pro image generation for study and research visuals |
| Search AI Mode (US) | Google AI Pro and Ultra subscribers | Pro images inside Search results |
| Vertex AI | Enterprises | Scaled production use through Google Cloud |
If you are building an app, read our Google AI Studio beginner guide next — it walks through API keys and your first project. If you work in an IDE, the Google Antigravity guide covers the agent-based editor where Pro generates layouts and mockups.
How to Make Your First Image in the Gemini App
Your first image takes about two minutes. Follow the steps exactly, and spend most of your effort on the prompt — the free quota is small, so a good first prompt saves you a wasted generation.
- Open the Gemini app. Go to gemini.google.com on the web, or open the Gemini app on your phone, and sign in.
- Pick the Thinking model. Tap the model dropdown above the prompt box — it may say Fast or 2.5 Flash — and select Thinking. This switches you to Gemini 3 Pro, which is what powers Nano Banana Pro.
- Choose Create images. This tells Gemini you want an image, not a chat answer.
- Write your prompt. Start with "create an image..." and then describe the subject, the setting, the lighting, and the style in one sentence.
- Set the size. Pick your aspect ratio and resolution before you generate: 1K for a quick check, 2K for general use, 4K for print or text-heavy work.
- Tap Generate and review. Look at the result closely, then change one thing at a time — a color, a word of text, a background — and generate again.
Here is a first prompt that works well as a test. It gives the model a subject, a setting, lighting, and an exact piece of text, which is exactly what Pro is good at:
Example prompt: "Create a 16:9 poster of a small bakery at sunrise, warm golden light through the front window, a chalkboard menu that reads 'Fresh Bread', photorealistic style, cozy and inviting."
New to Gemini itself? Our Gemini beginner guide covers the app's models, the free-vs-paid line, and the settings to check first.
How to Use Nano Banana Pro Reference Images
Nano Banana Pro reference images are the model's biggest advantage over the original. You can upload up to 14 images in a single prompt, and the model keeps up to 5 people's appearance consistent across the result. That is what makes it useful for product photos, character sheets, and brand mockups instead of one-off pictures.
Upload references that do different jobs. A product shot tells the model what the thing looks like. A face photo tells it who the person is. A style image tells it the mood, colors, or art direction. An environment photo tells it where the scene happens. Tell the model which is which in your prompt, for example: "Use the first image as the product, the second image as the background, and keep the person's face from the third image."
Reference images work best for four jobs. First, product campaigns: the same bottle, shoe, or gadget in different scenes without reshooting it. Second, character consistency: the same person or mascot across a series of images. Third, style transfer: your sketch, painting, or brand palette applied to a new image. Fourth, scene building: combining separate elements — a gown, plants, and a chair, in Google's own example — into one finished photograph.
Two rules to follow. Only upload images you have the right to use, and never upload someone else's face to create images of them without their permission. Also, check the result against your reference: if a logo or a face drifts, say so in your next prompt ("keep the logo exactly as in the reference image") rather than starting over from scratch.
How to Pick the Right Resolution and Aspect Ratio
Nano Banana Pro offers 1K, 2K, and 4K output, and the right choice depends on where the image will live. Higher resolution is slower and costs more on the API, so do not default to 4K for everything.
| Resolution | Use it for | Why |
|---|---|---|
| 1K | Drafts, quick checks, social posts | Fastest and cheapest; good enough to judge a composition |
| 2K | Blog images, slides, most marketing use | The sweet spot — sharp on screens without the 4K wait |
| 4K | Print, posters, text-heavy infographics | Fine text stays legible and details survive printing |
Aspect ratio matters just as much. Use 1:1 for Instagram posts and profile images, 16:9 for video thumbnails, slides, and blog headers, and 9:16 for Reels, TikTok, and Shorts. Nano Banana Pro 4K output is the choice for print and posters, where fine text must stay legible at full size. State the ratio in your prompt as well as in the settings — "create a 9:16 image" — because the model plans the composition around it, and a square image cropped to vertical later will lose the edges you wanted.
How to Edit an Image With a Text Prompt
Nano Banana Pro edits images the same way it creates them: you describe the change in plain language. Upload the image, then tell the model what to change. The model can refine a selected part of an image, change the camera angle, shift the focus, re-grade the colors, or change the lighting — for example, turning day into night or adding a soft bokeh effect.
Good edit prompts are specific about one thing. "Turn this photo into a watercolor painting" works. "Add a red umbrella over the woman" works. "Change the lighting to golden hour and add soft fog" works. "Focus on the flowers in the foreground" works. Vague prompts like "make it better" do not, because the model has no way to know what better means to you.
One creative trick worth trying: draw on your uploaded image first with colored markers to sketch where things should go, then prompt the model to interpret your doodles. It is a fast way to show a layout — a product here, text there — without writing a long paragraph. Then refine with a follow-up prompt that names the one detail that is still wrong.
Nano Banana Pro Free: What the Free Quota Actually Gets You
The free quota is small on purpose. Free users in the Gemini app get a limited number of Nano Banana Pro generations — Wondershare's Filmora guide puts it at around 2 images per day — and once that quota is used up, the app quietly switches you back to the original Nano Banana model. Google AI Plus, Pro, and Ultra subscribers get higher quotas.
| Plan | Pro quota | Visible watermark |
|---|---|---|
| Free | Limited — around 2 images a day, then you fall back to the original Nano Banana | Yes, the Gemini sparkle |
| Google AI Plus / Pro | Higher quotas than free | Yes, the Gemini sparkle |
| Google AI Ultra | Highest quotas, plus Flow access | No visible watermark |
| Google AI Studio (API) | Pay per image, no quota in the same sense | No visible watermark |
Two things about the watermark are worth knowing before you plan any work around it. Every image Google's tools generate carries an invisible SynthID watermark, and you can upload an image to the Gemini app and ask whether Google AI made it. On top of that, free and Pro-tier users also get a visible Gemini sparkle watermark on their images. Ultra subscribers and Google AI Studio users do not get the visible watermark, because Google wants them to have a clean canvas for professional work.
If you are paying per image through the API, the cost is easy to work out. Google's Gemini API pricing, as checked on July 23, 2026, lists Nano Banana Pro output at $0.134 per 1K or 2K image and $0.24 per 4K image. That means 10 images at 2K cost $1.34, while 10 at 4K cost $2.40 — nearly double for the extra resolution. Check the current price on Google's API pricing page before you budget a large batch, because API prices do change.
Nano Banana Pro vs Nano Banana: Which One Should You Use?
Use the original Nano Banana when you are exploring. Use Nano Banana Pro when the image has to ship. That is the whole decision, and the table below makes it concrete.
| Ask yourself | Use the original Nano Banana | Use Nano Banana Pro |
|---|---|---|
| Am I still deciding what I want? | Yes — fast iterations cost little | No — save Pro for the final direction |
| Does the image contain text? | Short labels only | Yes — posters, menus, infographics, mockups |
| Must a person or product stay identical? | No | Yes — up to 14 references, 5 people |
| Is the fact in the image real? | No | Yes — Search grounding checks facts first |
| Is this going to print? | No | Yes — 4K output |
| Am I in a hurry? | Yes — it is the faster model | No — Pro is slower by design |
A practical workflow is to draft with the original Nano Banana, pick the winning direction, and then rebuild that one image in Pro with your reference images attached. You get the speed of the fast model for exploring and the quality of Pro for the image people actually see.
How to Use Nano Banana Pro as a Developer
Developers do not use the Gemini app at all. Nano Banana Pro is available as Gemini 3 Pro Image through the Gemini API and Google AI Studio, with the model ID gemini-3-pro-image. Enterprises can build with it in Vertex AI, and it is also available in the Antigravity IDE for layouts and mockups.
The developer route is the right one when you need images inside an app, a repeatable pipeline, or a batch of assets with the same style. The app route is the right one when you are making images by hand. If you are weighing the API against a subscription, compare the per-image price above with your monthly volume: at $0.134 per 2K image, 100 images cost $13.40 through the API, while a subscription buys you a quota plus the rest of Google's AI bundle.
5 Prompting Tips That Work With Nano Banana Pro
Prompting Pro is closer to writing a creative brief than chatting. The model reasons about your words before it draws, so clear instructions pay off more here than in the original model.
1. Put subject, lighting, and style in one sentence. "A cinematic cityscape at sunset, soft bokeh lighting, pastel cyberpunk style" gives the model everything it needs to plan. Long, rambling paragraphs with three different styles in them do not.
2. Describe how things relate to each other. "A small dog balancing on a surfboard beneath a neon sign at night" works far better than "dog, surfboard, sign," because the model understands positions and relationships, not just a list of objects.
3. Put exact text in quotes. If a poster must say "Fresh Bread," write the words in quotes in your prompt. Short, quoted text renders far more reliably than a long paragraph the model has to summarize into an image.
4. Change one thing per generation. When a result is almost right, fix the single wrong detail — the background, one word, the lighting — and leave the rest alone. Changing five things at once makes it impossible to tell which change helped.
5. Use Search-grounded prompts for factual images. Pro can pull real-world facts into infographics, diagrams, maps, and weather visuals. Ask for the fact you want shown — a recipe, a timeline, a sports result — and then verify the numbers yourself before you publish, because the model can still get a detail wrong.
Common Problems and How to Fix Them
Most Nano Banana Pro problems come from the model selection or the quota, not from the prompt. Check these first before you rewrite anything.
It generated with the wrong model. You probably left the Fast model selected. Open the model dropdown and choose Thinking before you type your prompt. If your image comes back fast and rough, that is the original Nano Banana, not Pro.
The text in my image is misspelled. Shorten the text, put it in quotes, and try again. Long paragraphs of text are the hardest thing for any image model. If it still fails, generate the image without text and add the text in an editor.
I ran out of free images. Your daily Pro quota is spent and the app has switched you to the original Nano Banana. Wait until the quota resets, or move to a paid plan if you need Pro every day.
There is a watermark on my image. Free and Pro-tier images carry the visible Gemini sparkle watermark. Only Ultra subscribers and Google AI Studio outputs skip it. The invisible SynthID watermark is always present and cannot be removed.
Generation is slow. That is normal for Pro — the model reasons through your prompt before rendering. Use the original Nano Banana for quick drafts and save Pro for the final version.
Once you have your finished image, you can turn it into a video. Our CapCut beginner guide shows how to cut stills into a short video with captions and a clean export.
Sources
- Google — "Introducing Nano Banana Pro" (blog.google, November 20, 2025; updated March 19, 2026): launch details, the Gemini 3 Pro Image model, up to 14 reference images and 5 people, 2K and 4K output, SynthID and visible watermark rules, access routes, and free-tier quotas.
- Wondershare Filmora — "How to Use Nano Banana Pro for Free? Complete Beginner's Guide [2026]" (filmora.wondershare.com): Gemini app steps, the Thinking model selection, the free limit of around 2 images per day, and the fallback to the original Nano Banana.
- GlobalGPT — "How to Use Nano Banana Pro: Complete Beginner's Guide" (glbgpt.com): first-image steps, reference-image guidance, the gemini-3-pro-image model ID, and API pricing of $0.134 per 1K/2K image and $0.24 per 4K image, checked July 23, 2026.
- PBC Today — "Vectorworks 2027's AI Visualizer with Google's Nano Banana" (pbctoday.co.uk, October 27, 2026): third-party tools listing Nano Banana 2 and Nano Banana Pro as separate models.
Frequently Asked Questions
Is Nano Banana Pro free?
Partly. Free users can try Nano Banana Pro in the Gemini app with a limited quota — around 2 images per day — and after that the app switches back to the original Nano Banana model. Google AI Plus, Pro, and Ultra subscribers get higher quotas, and developers pay per image through the API.
What is the difference between Nano Banana and Nano Banana Pro?
The original Nano Banana is the Gemini 2.5 Flash Image model: fast and good for quick edits and experiments. Nano Banana Pro is the Gemini 3 Pro Image model: it reasons about your prompt, blends up to 14 reference images, keeps up to 5 people consistent, renders accurate text, and outputs up to 4K. It is slower and more expensive, but the results hold up for final work.
How many reference images can Nano Banana Pro use?
Up to 14 reference images in a single prompt, with up to 5 people's appearance kept consistent. Use them for different jobs — product shots, faces, style references, and environments — and tell the model in your prompt which image plays which role.
Does Nano Banana Pro add a watermark to images?
Yes, in two ways. Every image carries an invisible SynthID watermark, and you can ask the Gemini app whether an image was made by Google AI. Free and Google AI Pro users also get a visible Gemini sparkle watermark on their images. Ultra subscribers and Google AI Studio outputs do not carry the visible watermark.
What is the Nano Banana Pro API model name, and how much does it cost?
The API model ID is gemini-3-pro-image, available through the Gemini API and Google AI Studio. Pricing checked on July 23, 2026 lists $0.134 per 1K or 2K image and $0.24 per 4K image, so 10 images at 2K cost $1.34 and 10 at 4K cost $2.40. Check Google's current API pricing page before budgeting a large batch.