How to Use Nano Banana 2 API: Complete Tutorial with Code Examples (2026)

TL;DR

Nano Banana 2 (gemini-3.1-flash-image-preview) launched Feb 26, 2026 — Pro-level quality at Flash speed
Requires a paid API key. Free tier quota for image generation is zero
Google pricing: $0.101/image (2K), $0.150/image (4K). GPTImage2 offers ~20% lower at $0.0806 (2K)
Three ways to access: Google AI Studio (no code), Gemini API (Python/Node.js), or via a unified gateway
This tutorial gets you generating images in under 10 minutes

What Is Nano Banana 2?

Nano Banana 2 is Google's latest image generation model, built on Gemini 3.1 Flash Image. On Feb 26, 2026, Nano Banana 2 replaced Nano Banana Pro as the default image model across Gemini app's Fast, Thinking, and Pro modes.

Don't confuse it with Nano Banana Pro (gemini-3-pro-image-preview), which is a different model optimized for maximum fidelity. NB2 is the speed-and-cost play — near-Pro quality, significantly faster.

Key specs (source: Google blog):

Model ID: gemini-3.1-flash-image-preview
Resolution: 512px to 4K, native aspect ratios including 4:1, 1:4, 8:1, 1:8
Subject consistency: Up to 5 characters + 14 objects per generation
Text rendering: Improved multi-language in-image text
Thinking levels: Minimal (default) vs. High/Dynamic for complex prompts
AI identification: SynthID + C2PA Content Credentials

For a detailed NB2 vs Pro comparison, see our head-to-head breakdown.

Prerequisites: What You Need Before Starting

This is where most developers get stuck. On Reddit, the #1 complaint is hitting quota errors without understanding why.

The answer is simple: image generation requires a paid API key. No exceptions.

If you try with a free key, you'll get:

"Quota exceeded for metric: generativelanguage.googleapis.com/generate_content_free_tier_input_token_count, limit: 0"

Option A: Google Cloud Billing

Go to Google AI Studio
Click Get API Key → Create API key
Link a billing account in Google Cloud Console → Billing
Enable the Gemini API in your project
Your key now works for image generation

Option B: Third-Party Gateway

If you'd rather skip Google Cloud billing setup, gateways like GPTImage2 route to the same model with simpler onboarding — sign up, get a key, add credits, start generating.

Method 1: Google AI Studio (No Code)

The fastest way to test without writing code:

Open AI Studio
Select gemini-3.1-flash-image-preview from the model dropdown
Type a prompt: "A photorealistic golden retriever puppy in a sunlit meadow"
Click Run

If you see "only available for paid tier users," your billing isn't linked yet (see Prerequisites above).

Great for prompt testing before you write any integration code.

Method 2: Gemini API Direct

Python

pip install google-generativeai

import google.generativeai as genai
import base64
genai.configure(api_key="YOUR_PAID_API_KEY")
model = genai.GenerativeModel("gemini-3.1-flash-image-preview")
response = model.generate_content(

"A photorealistic golden retriever puppy sitting in a sunlit meadow, "

"soft bokeh background, warm afternoon light",

generation_config=genai.GenerationConfig(

response_modalities=["image", "text"],  # Optional; model defaults to both text and image

),

)
for part in response.parts:

if part.inline_data:

image_data = base64.b64decode(part.inline_data.data)

with open("output.png", "wb") as f:

f.write(image_data)

print("Image saved to output.png")

elif part.text:

print(f"Response: {part.text}")

Node.js

npm install @google/generative-ai

const { GoogleGenerativeAI } = require("@google/generative-ai");
const fs = require("fs");
const genAI = new GoogleGenerativeAI("YOUR_PAID_API_KEY");
async function generateImage() {

const model = genAI.getGenerativeModel({

model: "gemini-3.1-flash-image-preview",

});
const result = await model.generateContent({

contents: [

{

role: "user",

parts: [

{

text: "A photorealistic golden retriever puppy sitting in a sunlit meadow, soft bokeh background, warm afternoon light",

},

],

},

],

generationConfig: {

responseModalities: ["image", "text"],

},

});
const response = result.response;

for (const part of response.candidates[0].content.parts) {

if (part.inlineData) {

const buffer = Buffer.from(part.inlineData.data, "base64");

fs.writeFileSync("output.png", buffer);

console.log("Image saved to output.png");

} else if (part.text) {

console.log("Response:", part.text);

}

}

}
generateImage();

Method 3: OpenAI-Compatible Gateway (GPTImage2)

If you already use the OpenAI SDK, you can generate images without changing your code structure — just point to a different base URL.

Python (OpenAI SDK)

pip install openai

from openai import OpenAI
client = OpenAI(

api_key="YOUR_GPTIMAGE2_API_KEY",

base_url="https://api.gptimage2/v1&quot;,

)
response = client.chat.completions.create(

model="gemini-3.1-flash-image-preview",

messages=[

{

"role": "user",

"content": "A photorealistic golden retriever puppy sitting in a sunlit meadow",

}

],

)
print(response.choices[0].message.content)

Node.js (OpenAI SDK)

npm install openai

const OpenAI = require("openai");
const client = new OpenAI({

apiKey: "YOUR_GPTIMAGE2_API_KEY",

baseURL: "https://api.gptimage2/v1&quot;,

});
async function generateImage() {

const response = await client.chat.completions.create({

model: "gemini-3.1-flash-image-preview",

messages: [

{

role: "user",

content:

"A photorealistic golden retriever puppy sitting in a sunlit meadow",

},

],

});
console.log(response.choices[0].message.content);

}
generateImage();

Why use a gateway? One API key for multiple models (NB2, Pro, GPT-4o Image, Seedream). No Google Cloud billing setup. Same model, same output.

Pricing Breakdown

Resolution	Google Direct	GPTImage2	Savings
2K	$0.101	$0.0806	~20%
4K	$0.150	$0.1210	~19%

Text token costs are negligible (~$0.0001 per prompt). The image output token is where the cost is.

For batch workloads (100+ images/day), the savings compound quickly.

Prompt Engineering Tips

Be Specific

❌ "A dog"
✅ "A photorealistic golden retriever puppy sitting in a sunlit meadow, soft bokeh background, warm afternoon light"

Use Style Keywords

Photography styles: photorealistic, cinematic, studio lighting, 35mm film

Art styles: watercolor, oil painting, digital art, anime, pixel art

Leverage Thinking Levels

For complex prompts with multiple subjects, enable higher thinking:

generation_config=genai.GenerationConfig(
    response_modalities=["image", "text"],
    thinking_config={"thinking_budget": 1024},  # Higher = more reasoning
)

Multi-Turn Editing

Generate an image, then refine it in follow-up messages:

Turn 1: "A modern minimalist living room with a large window"
Turn 2: "Add a orange tabby cat sleeping on the couch"
Turn 3: "Change the time to sunset, warm golden light through the window"

Troubleshooting

Quota Exceeded Error

"Quota exceeded for metric: generate_content_free_tier_input_token_count"

Fix: You're using a free API key. Image generation requires billing enabled. See Prerequisites.

400 Bad Request

Common causes:

Typo in model ID
Unsupported aspect ratio

Text Response Instead of Image

responseModalities is optional. The model defaults to returning both text and image. If your SDK/client returns text-only, set responseModalities to include IMAGE:

generation_config=genai.GenerationConfig(
    response_modalities=["image", "text"],
)

FAQ

Q: Is Nano Banana 2 free to use via API?

A: No. Image generation requires a paid API key with billing enabled. Free tier quota is 0.

Q: What's the model ID for Nano Banana 2?

A: gemini-3.1-flash-image-preview. Don't confuse with gemini-3-pro-image-preview (Pro) or gemini-2.5-flash-image (original).

Q: How much does one image cost?

A: 2K: $0.101 Google direct. 4K: $0.150. Plus negligible text token costs (~$0.0001/prompt). Pricing as of Feb 2026.

Q: NB2 vs NB Pro — which should I use?

A: Default to NB2. Faster, ~50% cheaper. Use Pro only for maximum fidelity or precise text rendering (94% accuracy). See NB2 vs Pro comparison.

Q: What are the rate limits?

A: Google's RPM limits vary by billing tier and aren't fully documented. Implement exponential backoff for 429 errors.

Q: Can I edit images with NB2?

A: Yes. Multi-turn conversations — generate an image, then describe edits in follow-up messages.

Q: What resolutions are supported?

A: 512px, 1K, 2K, 4K. Aspect ratios: 1:1, 16:9, 9:16, 4:3, 3:4, 4:1, 1:4, 8:1, 1:8.

Q: Are images watermarked?

A: SynthID (invisible) + C2PA Content Credentials. Not visually visible but detectable by verification tools.

Q: How to switch from Pro to NB2?

A: Change model from gemini-3-pro-image-preview to gemini-3.1-flash-image-preview. API interface identical, no other code changes.

Start Generating

Three steps:

Get a paid API key (Google AI Studio or gptimage2/signup)
Copy the code above
Run it

→ GPTImage2 API docs → NB2 vs Pro comparison → Nano Banana 2 product page