Gemini API Image Generation: Complete Developer Guide (2026)

Generating Images with the Gemini API

The Gemini API gives developers programmatic access to Google's image generation models. Whether you're building a content tool, automating social media, or creating a product that needs AI-generated visuals, the Gemini API is one of the most capable and cost-effective options available.

Getting Started

Go to aistudio.google.com → API keys → Create API key. It's free to start with generous rate limits.

npm install @google/generative-ai

import { GoogleGenerativeAI } from "@google/generative-ai";

const genAI = new GoogleGenerativeAI(process.env.GEMINI_API_KEY);

async function generateImage(prompt) {
  const model = genAI.getGenerativeModel({
    model: "gemini-2.0-flash-exp-image-generation"
  });

  const response = await model.generateContent({
    contents: [{
      role: "user",
      parts: [{ text: prompt }]
    }],
    generationConfig: {
      responseModalities: ["image", "text"]
    }
  });

  const imagePart = response.response.candidates[0]
    .content.parts.find(p => p.inlineData);

  return imagePart.inlineData; // { mimeType, data (base64) }
}

Available Models for Image Generation

Handling Watermarks Programmatically

All Gemini API image outputs contain the same watermark as the web interface. For applications that need clean images, you have two options:

For low-volume use cases, process images through ilovewatermark.com/tool after generation.

const Jimp = require('jimp');

async function removeGeminiWatermark(imageBuffer) {
  const image = await Jimp.read(imageBuffer);
  const w = image.bitmap.width;
  const h = image.bitmap.height;

  // Load appropriate alpha mask
  const isLarge = w > 1024 && h > 1024;
  const maskSize = isLarge ? 96 : 48;
  const margin = isLarge ? 64 : 32;
  const maskPath = isLarge ? './bg_96.png' : './bg_48.png';

  const mask = await Jimp.read(maskPath);

  // Watermark position
  const wx = w - maskSize - margin;
  const wy = h - maskSize - margin;

  // Reverse alpha blending
  for (let row = 0; row < maskSize; row++) {
    for (let col = 0; col < maskSize; col++) {
      const maskPixel = Jimp.intToRGBA(mask.getPixelColor(col, row));
      const alpha = Math.max(maskPixel.r, maskPixel.g, maskPixel.b) / 255;

      if (alpha < 0.002) continue;
      const clampedAlpha = Math.min(alpha, 0.99);

      const imgPixel = Jimp.intToRGBA(
        image.getPixelColor(wx + col, wy + row)
      );

      const r = Math.max(0, Math.min(255,
        Math.round((imgPixel.r - clampedAlpha * 255) / (1 - clampedAlpha))
      ));
      const g = Math.max(0, Math.min(255,
        Math.round((imgPixel.g - clampedAlpha * 255) / (1 - clampedAlpha))
      ));
      const b = Math.max(0, Math.min(255,
        Math.round((imgPixel.b - clampedAlpha * 255) / (1 - clampedAlpha))
      ));

      image.setPixelColor(
        Jimp.rgbaToInt(r, g, b, 255),
        wx + col,
        wy + row
      );
    }
  }

  return await image.getBufferAsync(Jimp.MIME_PNG);
}

You'll need the bg_48.png and bg_96.png alpha mask files. These are available from the open-source gemini-watermark-remover project on GitHub.

Rate Limits and Pricing

Gemini API pricing for image generation (2026):

• Free tier: 15 requests per minute, 1,500 per day
• Pay-as-you-go: varies by model

For most content automation use cases, the free tier is sufficient. For high-volume production use, the pay-as-you-go pricing is very competitive.

Best Practices for Production

Ready to remove your Gemini watermark?

Free tool — no signup required. First 2 images free.