Integrate Gemini Nano Banana in Local Apps

Gemini 2.5 Flash Image API Guide (2026)

If you want to call Nano Banana (Gemini image generation) from a local mini app or desktop app, you need to configure API access in Google AI Studio and send requests in the correct JSON format.

This guide follows the latest Gemini API usage pattern and gives you a practical integration template.

1. Access and Base Configuration

Before coding, prepare your API key and endpoint.

Get API Key

Go to Google AI Studio
Sign in
Click Create API Key

Endpoint

https://generativelanguage.googleapis.com/v1beta/models/gemini-2.5-flash-image:generateContent

Request Method

POST

Required Headers

Content-Type: application/json
x-goog-api-key: YOUR_API_KEY

2. Request Format (JSON Payload)

Put your image prompt inside contents[].parts[].text.

To force image output, set response_modalities to ["IMAGE"].

{
  "contents": [
    {
      "parts": [
        { "text": "A cat drinking coffee on a cyberpunk rooftop, cinematic lighting, 4k resolution" }
      ]
    }
  ],
  "generationConfig": {
    "response_modalities": ["IMAGE"],
    "candidate_count": 1,
    "aspect_ratio": "1:1"
  }
}

3. Response Handling

Gemini image output is returned as Base64 in inline_data.data.

Response Shape (simplified)

{
  "candidates": [
    {
      "content": {
        "parts": [
          {
            "inline_data": {
              "mime_type": "image/png",
              "data": "iVBORw0KGgoAAAANSUhEUgAA..."
            }
          }
        ]
      }
    }
  ]
}

Mini-app Frontend Rendering

Build a data URL:

data:image/png;base64,<BASE64_STRING>

Then bind it to your <image> component.

4. Model Selection Tips

Choose model IDs based on latency, quality, and reasoning needs.

Nickname	Model ID	Typical Use
Nano Banana	`gemini-2.5-flash-image`	Fastest response, low-latency generation for mini apps.
Nano Banana 2	`gemini-3.1-flash-image-preview`	Better quality/understanding if available in your account.
Nano Banana Pro	`gemini-3-pro-image-preview`	Highest quality and advanced instruction following.

Note: preview model availability can vary by region/account and may change over time.

5. Quick cURL Test

Use this in early development to validate credentials and payload quickly:

curl "https://generativelanguage.googleapis.com/v1beta/models/gemini-2.5-flash-image:generateContent?key=YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -X POST \
  -d '{
    "contents": [
      { "parts": [ { "text": "A futuristic laboratory with neon lights" } ] }
    ],
    "generationConfig": {
      "response_modalities": ["IMAGE"]
    }
  }'

6. Security and Production Advice

Do not expose your API key directly in client-side code.

Recommended architecture:

Mini app / local app calls your own backend proxy (Node.js or Python)
Backend injects API key server-side
Backend calls Gemini API and returns safe response

Also plan for:

Rate limiting and retries
Request logging
Error mapping for frontend UX

7. Common Integration Checklist

API key created and enabled
Correct endpoint and model ID
x-goog-api-key in headers
response_modalities: ["IMAGE"] set
inline_data.data parsed correctly
Base64 data URL rendered in UI
API key protected by backend proxy

With this structure, you can quickly integrate Nano Banana image generation into local mini apps while keeping latency low and credentials secure.