Gemini Image Generation

How we generate images via the Gemini API, ported from the getcirclesorg repo.

Setup

Install the SDK:

pnpm add -D @google/genai

Set the environment variable:

GOOGLE_GENAI_API_KEY=<your-key>

Basic usage

import { GoogleGenAI } from "@google/genai";

const ai = new GoogleGenAI({ apiKey: process.env.GOOGLE_GENAI_API_KEY });

const response = await ai.models.generateContent({
  model: "gemini-3-pro-image-preview",
  contents: "your prompt here",
  config: {
    responseModalities: ["Image"],
    imageConfig: {
      aspectRatio: "16:9",
    },
  },
});

// Extract the image from the response
const part = response.candidates[0].content.parts.find((p) => p.inlineData);
const imageBuffer = Buffer.from(part.inlineData.data, "base64");

Available models

Image config options

aspectRatio

"1:1", "1:4", "1:8", "2:3", "3:2", "3:4", "4:1", "4:3", "4:5", "5:4", "8:1", "9:16", "16:9", "21:9"

Aspect ratio is not part of the prompt -- it's a config parameter.

imageSize

"512" (Flash only), "1K" (default), "2K", "4K" -- must use uppercase K.

Response modalities

• ["Image"] -- image-only response
• ["TEXT", "IMAGE"] -- model can return text alongside the image

Post-processing with sharp

The getcirclesorg scripts use sharp to resize and compress after generation:

import sharp from "sharp";

await sharp(imageBuffer)
  .resize(600, 400, { fit: "cover" })
  .jpeg({ quality: 85 })
  .toFile("output.jpg");

Our prompts

See image-prompts.md for the full set of prompts and design tokens for DeepSpace/TracePlot assets.