How to use gemini-2.5-flash-image with Google AI Studio

Here’s the fastest way to get started with gemini-2.5-flash-image (Google’s native text-to-image & image-editing model).

1) Pick your access path

• Gemini API (AI Studio) — simplest way; auth via API key. Great for apps, servers, and quick prototyping. (Google AI for Developers)

• Vertex AI — GCP-managed, quotas/billing/enterprise controls; same model name. Note: preview image models are being retired; migrate to gemini-2.5-flash-image. (Google Cloud)

---

2) Get an API key (AI Studio)

Create a key in Google AI Studio → “Get API key,” then set GEMINI_API_KEY in your environment. Docs + model card: (Google AI for Developers)

---

3) Minimal “text → image” call

cURL (Gemini API)

curl -s -X POST \
  "https://generativelanguage.googleapis.com/v1beta/models/gemini-2.5-flash-image:generateContent" \
  -H "x-goog-api-key: $GEMINI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "contents":[{"parts":[{"text":
      "A cozy, cinematic coffee shop interior at golden hour, shallow depth of field, 35mm, soft rim light"}
    ]}]
  }'

The response includes an image as base64 in candidates[0].content.parts[].inlineData.data. (Google AI for Developers)

JavaScript (Node)

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

const ai = new GoogleGenerativeAI(process.env.GEMINI_API_KEY);
const prompt = "Minimalist poster of a red fox in the snow, high contrast, grainy film look";

const res = await ai.models.generateContent({
  model: "gemini-2.5-flash-image",
  contents: prompt,
});
for (const part of res.candidates[0].content.parts) {
  if (part.inlineData) {
    fs.writeFileSync("output.png", Buffer.from(part.inlineData.data, "base64"));
  }
}

(Google AI for Developers)

Python

from google import genai
from google.genai import types
from io import BytesIO
from PIL import Image

client = genai.Client()  # reads GOOGLE_API_KEY or similar env var
prompt = "Architectural render of a tranquil Japanese courtyard, dusk, volumetric light"

resp = client.models.generate_content(
    model="gemini-2.5-flash-image",
    contents=prompt,
)
image_bytes = next(
    (p.inline_data.data for p in resp.candidates[0].content.parts if getattr(p, "inline_data", None)),
    None,
)
if image_bytes:
    img = Image.open(BytesIO(image_bytes))
    img.save("courtyard.png")

(Google AI for Developers)

---

4) Control output (aspect ratio, image-only, etc.)

Add a config to your call:

JS (force image-only + 16:9)

const res = await ai.models.generateContent({
  model: "gemini-2.5-flash-image",
  contents: "Wide cinematic landscape of Icelandic black sand beach, long exposure",
  config: {
    responseModalities: ["Image"],
    imageConfig: { aspectRatio: "16:9" }
  }
});

(Also available in Python/Go.) (Google AI for Developers)

---

5) Image editing / multi-image fusion

Pass input images (base64) alongside your text instruction.

cURL (edit an image)

IMG_BASE64=$(base64 -w0 input.jpg)
curl -s -X POST \
  "https://generativelanguage.googleapis.com/v1beta/models/gemini-2.5-flash-image:generateContent" \
  -H "x-goog-api-key: $GEMINI_API_KEY" -H 'Content-Type: application/json' \
  -d "{
    \"contents\": [{
      \"parts\": [
        {\"text\": \"Replace the sky with dramatic storm clouds; keep colors natural\"},
        {\"inline_data\": {\"mime_type\":\"image/jpeg\",\"data\":\"$IMG_BASE64\"}}
      ]
    }]
  }"

You can include multiple inline_data parts to merge images (e.g., place a product into a new scene). (Google AI for Developers)

---

6) Prompting tips (huge quality boost)

Describe a full scene (camera, lens, lighting, mood, materials). Narrative prompts reliably beat keyword lists. The official guide includes templates for photorealistic, logo, sticker, and negative-space compositions. (Google AI for Developers)

---

7) Pricing, safety & watermarks

• Pricing (Gemini API): blog post lists $30 / 1M output tokens; each image ~1290 output tokens (~$0.039/image). (Check your region/account for current pricing.) (Google Developers Blog)

• Retirements: preview image models (gemini-2.0-flash-preview-image-generation, gemini-2.5-flash-image-preview) retire Oct 31, 2025—use gemini-2.5-flash-image. (Google Cloud)

• Watermarking: Images include SynthID invisible watermark for provenance. (Google Developers Blog)

---

8) Vertex AI variant (if you’re on GCP)

Use the Vertex AI SDK or REST with the same model name gemini-2.5-flash-image; samples and quickstart in Google Cloud docs. (Google Cloud)

---

9) Common gotchas

• No image file? Ensure you’re reading inlineData.data (base64) from the first candidate and writing it as bytes. (Google AI for Developers)

• Only images back (no text): set responseModalities: ['Image']. (Google AI for Developers)

• Sizes/ratios: use imageConfig.aspectRatio (e.g., "1:1", "4:5", "16:9"). (Google AI for Developers)

If you tell me your stack (Node, Python, or Vertex AI) and your use case (new image vs edit vs fusion), I’ll tailor a drop-in snippet for your app.