Generate images

The Gemini API supports image generation using Gemini 2.0 Flash Experimental and using Imagen 3. This guide helps you get started with both models.

For image prompting guidance, check out the Imagen prompt guide section.

Before you begin

Before calling the Gemini API, ensure you have your SDK of choice installed, and a Gemini API key configured and ready to use.

Generate images using Gemini

Gemini 2.0 Flash Experimental supports the ability to output text and inline images. This lets you use Gemini to conversationally edit images or generate outputs with interwoven text (for example, generating a blog post with text and images in a single turn). All generated images include a SynthID watermark, and images in Google AI Studio include a visible watermark as well.

The following example shows how to use Gemini 2.0 to generate text-and-image output:

from google import genai
from google.genai import types
from PIL import Image
from io import BytesIO
import base64

client = genai.Client()

contents = ('Hi, can you create a 3d rendered image of a pig '
            'with wings and a top hat flying over a happy '
            'futuristic scifi city with lots of greenery?')

response = client.models.generate_content(
    model="gemini-2.0-flash-exp-image-generation",
    contents=contents,
    config=types.GenerateContentConfig(
      response_modalities=['TEXT', 'IMAGE']
    )
)

for part in response.candidates[0].content.parts:
  if part.text is not None:
    print(part.text)
  elif part.inline_data is not None:
    image = Image.open(BytesIO((part.inline_data.data)))
    image.save('gemini-native-image.png')
    image.show()

import { GoogleGenAI, Modality } from "@google/genai";
import * as fs from "node:fs";

async function main() {

  const ai = new GoogleGenAI({ apiKey: "GEMINI_API_KEY" });

  const contents =
    "Hi, can you create a 3d rendered image of a pig " +
    "with wings and a top hat flying over a happy " +
    "futuristic scifi city with lots of greenery?";

  // Set responseModalities to include "Image" so the model can generate  an image
  const response = await ai.models.generateContent({
    model: "gemini-2.0-flash-exp-image-generation",
    contents: contents,
    config: {
      responseModalities: [Modality.TEXT, Modality.IMAGE],
    },
  });
  for (const part of response.candidates[0].content.parts) {
    // Based on the part type, either show the text or save the image
    if (part.text) {
      console.log(part.text);
    } else if (part.inlineData) {
      const imageData = part.inlineData.data;
      const buffer = Buffer.from(imageData, "base64");
      fs.writeFileSync("gemini-native-image.png", buffer);
      console.log("Image saved as gemini-native-image.png");
    }
  }
}

main();

curl -s -X POST \
  "https://generativelanguage.googleapis.com/v1beta/models/gemini-2.0-flash-exp-image-generation:generateContent?key=$GEMINI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "contents": [{
      "parts": [
        {"text": "Hi, can you create a 3d rendered image of a pig with wings and a top hat flying over a happy futuristic scifi city with lots of greenery?"}
      ]
    }],
    "generationConfig":{"responseModalities":["TEXT","IMAGE"]}
  }' \
  | grep -o '"data": "[^"]*"' \
  | cut -d'"' -f4 \
  | base64 --decode > gemini-native-image.png

Depending on the prompt and context, Gemini will generate content in different modes (text to image, text to image and text, etc.). Here are some examples:

• Text to image
  • Example prompt: "Generate an image of the Eiffel tower with fireworks in the background."
• Text to image(s) and text (interleaved)
  • Example prompt: "Generate an illustrated recipe for a paella."
• Image(s) and text to image(s) and text (interleaved)
  • Example prompt: (With an image of a furnished room) "What other color sofas would work in my space? can you update the image?"
• Image editing (text and image to image)
  • Example prompt: "Edit this image to make it look like a cartoon"
  • Example prompt: [image of a cat] + [image of a pillow] + "Create a cross stitch of my cat on this pillow."
• Multi-turn image editing (chat)
  • Example prompts: [upload an image of a blue car.] "Turn this car into a convertible." "Now change the color to yellow."

Image editing with Gemini

To perform image editing, add an image as input. The following example demonstrats uploading base64 encoded images. For multiple images and larger payloads, check the image input section.

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

import PIL.Image

image = PIL.Image.open('/path/to/image.png')

client = genai.Client()

text_input = ('Hi, This is a picture of me.'
            'Can you add a llama next to me?',)

response = client.models.generate_content(
    model="gemini-2.0-flash-exp-image-generation",
    contents=[text_input, image],
    config=types.GenerateContentConfig(
      response_modalities=['TEXT', 'IMAGE']
    )
)

for part in response.candidates[0].content.parts:
  if part.text is not None:
    print(part.text)
  elif part.inline_data is not None:
    image = Image.open(BytesIO(part.inline_data.data))
    image.show()

import { GoogleGenAI, Modality } from "@google/genai";
import * as fs from "node:fs";

async function main() {

  const ai = new GoogleGenAI({ apiKey: "GEMINI_API_KEY" });

  // Load the image from the local file system
  const imagePath = "path/to/image.png";
  const imageData = fs.readFileSync(imagePath);
  const base64Image = imageData.toString("base64");

  // Prepare the content parts
  const contents = [
    { text: "Can you add a llama next to the image?" },
    {
      inlineData: {
        mimeType: "image/png",
        data: base64Image,
      },
    },
  ];

  // Set responseModalities to include "Image" so the model can generate an image
  const response = await ai.models.generateContent({
    model: "gemini-2.0-flash-exp-image-generation",
    contents: contents,
    config: {
      responseModalities: [Modality.TEXT, Modality.IMAGE],
    },
  });
  for (const part of response.candidates[0].content.parts) {
    // Based on the part type, either show the text or save the image
    if (part.text) {
      console.log(part.text);
    } else if (part.inlineData) {
      const imageData = part.inlineData.data;
      const buffer = Buffer.from(imageData, "base64");
      fs.writeFileSync("gemini-native-image.png", buffer);
      console.log("Image saved as gemini-native-image.png");
    }
  }
}

main();

IMG_PATH=/path/to/your/image1.jpeg

if [[ "$(base64 --version 2>&1)" = *"FreeBSD"* ]]; then
  B64FLAGS="--input"
else
  B64FLAGS="-w0"
fi

IMG_BASE64=$(base64 "$B64FLAGS" "$IMG_PATH" 2>&1)

curl -X POST \
  "https://generativelanguage.googleapis.com/v1beta/models/gemini-2.0-flash-exp-image-generation:generateContent?key=$GEMINI_API_KEY" \
    -H 'Content-Type: application/json' \
    -d "{
      \"contents\": [{
        \"parts\":[
            {\"text\": \"'Hi, This is a picture of me. Can you add a llama next to me\"},
            {
              \"inline_data\": {
                \"mime_type\":\"image/jpeg\",
                \"data\": \"$IMG_BASE64\"
              }
            }
        ]
      }],
      \"generationConfig\": {\"responseModalities\": [\"TEXT\", \"IMAGE\"]}
    }"  \
  | grep -o '"data": "[^"]*"' \
  | cut -d'"' -f4 \
  | base64 --decode > gemini-edited-image.png

Limitations

• For best performance, use the following languages: EN, es-MX, ja-JP, zh-CN, hi-IN.
• Image generation does not support audio or video inputs.
• Image generation may not always trigger:
  • The model may output text only. Try asking for image outputs explicitly (e.g. "generate an image", "provide images as you go along", "update the image").
  • The model may stop generating partway through. Try again or try a different prompt.
• When generating text for an image, Gemini works best if you first generate the text and then ask for an image with the text.

Choose a model

Which model should you use to generate images? It depends on your use case.

Gemini 2.0 is best for producing contextually relevant images, blending text + images, incorporating world knowledge, and reasoning about images. You can use it to create accurate, contextually relevant visuals embedded in long text sequences. You can also edit images conversationally, using natural language, while maintaining context throughout the conversation.

If image quality is your top priority, then Imagen 3 is a better choice. Imagen 3 excels at photorealism, artistic detail, and specific artistic styles like impressionism or anime. Imagen 3 is also a good choice for specialized image editing tasks like updating product backgrounds, upscaling images, and infusing branding and style into visuals. You can use Imagen 3 to create logos or other branded product designs.

Generate images using Imagen 3

The Gemini API provides access to Imagen 3, Google's highest quality text-to-image model, featuring a number of new and improved capabilities. Imagen 3 can do the following:

• Generate images with better detail, richer lighting, and fewer distracting artifacts than previous models
• Understand prompts written in natural language
• Generate images in a wide range of formats and styles
• Render text more effectively than previous models

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

client = genai.Client(api_key='GEMINI_API_KEY')

response = client.models.generate_images(
    model='imagen-3.0-generate-002',
    prompt='Robot holding a red skateboard',
    config=types.GenerateImagesConfig(
        number_of_images= 4,
    )
)
for generated_image in response.generated_images:
  image = Image.open(BytesIO(generated_image.image.image_bytes))
  image.show()

import { GoogleGenAI } from "@google/genai";
import * as fs from "node:fs";

async function main() {

  const ai = new GoogleGenAI({ apiKey: "GEMINI_API_KEY" });

  const response = await ai.models.generateImages({
    model: 'imagen-3.0-generate-002',
    prompt: 'Robot holding a red skateboard',
    config: {
      numberOfImages: 4,
    },
  });

  let idx = 1;
  for (const generatedImage of response.generatedImages) {
    let imgBytes = generatedImage.image.imageBytes;
    const buffer = Buffer.from(imgBytes, "base64");
    fs.writeFileSync(`imagen-${idx}.png`, buffer);
    idx++;
  }
}

main();

curl -X POST \
    "https://generativelanguage.googleapis.com/v1beta/models/imagen-3.0-generate-002:predict?key=GEMINI_API_KEY" \
    -H "Content-Type: application/json" \
    -d '{
        "instances": [
          {
            "prompt": "Robot holding a red skateboard"
          }
        ],
        "parameters": {
          "sampleCount": 4
        }
      }'

Imagen supports English only prompts at this time and the following parameters:

Imagen model parameters

(Naming conventions vary by programming language.)

• numberOfImages: The number of images to generate, from 1 to 4 (inclusive). The default is 4.
• aspectRatio: Changes the aspect ratio of the generated image. Supported values are "1:1", "3:4", "4:3", "9:16", and "16:9". The default is "1:1".
• personGeneration: Allow the model to generate images of people. The following values are supported:
  • "DONT_ALLOW": Block generation of images of people.
  • "ALLOW_ADULT": Generate images of adults, but not children. This is the default.

Imagen prompt guide

This section of the Imagen guide shows you how modifying a text-to-image prompt can produce different results, along with examples of images you can create.

Prompt writing basics

A good prompt is descriptive and clear, and makes use of meaningful keywords and modifiers. Start by thinking of your subject, context, and style.

1. Subject: The first thing to think about with any prompt is the subject: the object, person, animal, or scenery you want an image of.

2. Context and background: Just as important is the background or context in which the subject will be placed. Try placing your subject in a variety of backgrounds. For example, a studio with a white background, outdoors, or indoor environments.

3. Style: Finally, add the style of image you want. Styles can be general (painting, photograph, sketches) or very specific (pastel painting, charcoal drawing, isometric 3D). You can also combine styles.

After you write a first version of your prompt, refine your prompt by adding more details until you get to the image that you want. Iteration is important. Start by establishing your core idea, and then refine and expand upon that core idea until the generated image is close to your vision.

Imagen 3 can transform your ideas into detailed images, whether your prompts are short or long and detailed. Refine your vision through iterative prompting, adding details until you achieve the perfect result.

Short prompts let you generate an image quickly.

Longer prompts let you add specific details and build your image.

Additional advice for Imagen prompt writing:

• Use descriptive language: Employ detailed adjectives and adverbs to paint a clear picture for Imagen 3.
• Provide context: If necessary, include background information to aid the AI's understanding.
• Reference specific artists or styles: If you have a particular aesthetic in mind, referencing specific artists or art movements can be helpful.
• Use prompt engineering tools: Consider exploring prompt engineering tools or resources to help you refine your prompts and achieve optimal results.
• Enhancing the facial details in your personal and group images:
  • Specify facial details as a focus of the photo (for example, use the word "portrait" in the prompt).

Generate text in images

Imagen can add text into images, opening up more creative image generation possibilities. Use the following guidance to get the most out of this feature:

• Iterate with confidence: You might have to regenerate images until you achieve the look you want. Imagen's text integration is still evolving, and sometimes multiple attempts yield the best results.
• Keep it short: Limit text to 25 characters or less for optimal generation.

• Multiple phrases: Experiment with two or three distinct phrases to provide additional information. Avoid exceeding three phrases for cleaner compositions.

  

• Guide Placement: While Imagen can attempt to position text as directed, expect occasional variations. This feature is continually improving.

• Inspire font style: Specify a general font style to subtly influence Imagen's choices. Don't rely on precise font replication, but expect creative interpretations.

• Font size: Specify a font size or a general indication of size (for example, small, medium, large) to influence the font size generation.

Prompt parameterization

To better control output results, you might find it helpful to parameterize the inputs into Imagen. For example, suppose you want your customers to be able to generate logos for their business, and you want to make sure logos are always generated on a solid color background. You also want to limit the options that the client can select from a menu.

In this example, you can create a parameterized prompt similar to the following:

A {logo_style} logo for a {company_area} company on a solid color background. Include the text {company_name}.

In your custom user interface, the customer can input the parameters using a menu, and their chosen value populates the prompt Imagen receives.

For example:

1. Prompt: A minimalist logo for a health care company on a solid color background. Include the text Journey.

   

2. Prompt: A modern logo for a software company on a solid color background. Include the text Silo.

   

3. Prompt: A traditional logo for a baking company on a solid color background. Include the text Seed.

   

Advanced prompt writing techniques

Use the following examples to create more specific prompts based on attributes like photography descriptors, shapes and materials, historical art movements, and image quality modifiers.

Photography

• Prompt includes: "A photo of..."

To use this style, start with using keywords that clearly tell Imagen that you're looking for a photograph. Start your prompts with "A photo of. . .". For example:

^{Image source: Each image was generated using its corresponding text prompt with the Imagen 3 model.}

Photography modifiers

In the following examples, you can see several photography-specific modifiers and parameters. You can combine multiple modifiers for more precise control.

1. Camera Proximity - Close up, taken from far away

2. Camera Position - aerial, from below

3. Lighting - natural, dramatic, warm, cold

4. Camera Settings - motion blur, soft focus, bokeh, portrait

5. Lens types - 35mm, 50mm, fisheye, wide angle, macro

6. Film types - black and white, polaroid

^{Image source: Each image was generated using its corresponding text prompt with the Imagen 3 model.}

Illustration and art

• Prompt includes: "A painting of...", "A sketch of..."

Art styles vary from monochrome styles like pencil sketches, to hyper-realistic digital art. For example, the following images use the same prompt with different styles:

"An [art style or creation technique] of an angular sporty electric sedan with skyscrapers in the background"

^{Image source: Each image was generated using its corresponding text prompt with the Imagen 2 model.}

Shapes and materials

• Prompt includes: "...made of...", "...in the shape of..."

One of the strengths of this technology is that you can create imagery that is otherwise difficult or impossible. For example, you can recreate your company logo in different materials and textures.

^{Image source: Each image was generated using its corresponding text prompt with the Imagen 3 model.}

Historical art references

• Prompt includes: "...in the style of..."

Certain styles have become iconic over the years. The following are some ideas of historical painting or art styles that you can try.

"generate an image in the style of [art period or movement] : a wind farm"

^{Image source: Each image was generated using its corresponding text prompt with the Imagen 3 model.}

Image quality modifiers

Certain keywords can let the model know that you're looking for a high-quality asset. Examples of quality modifiers include the following:

• General Modifiers - high-quality, beautiful, stylized
• Photos - 4K, HDR, Studio Photo
• Art, Illustration - by a professional, detailed

The following are a few examples of prompts without quality modifiers and the same prompt with quality modifiers.

^{Image source: Each image was generated using its corresponding text prompt with the Imagen 3 model.}

Aspect ratios

Imagen 3 image generation lets you set five distinct image aspect ratios.

1. Square (1:1, default) - A standard square photo. Common uses for this aspect ratio include social media posts.

2. Fullscreen (4:3) - This aspect ratio is commonly used in media or film. It is also the dimensions of most old (non-widescreen) TVs and medium format cameras. It captures more of the scene horizontally (compared to 1:1), making it a preferred aspect ratio for photography.

3. Portrait full screen (3:4) - This is the fullscreen aspect ratio rotated 90 degrees. This lets to capture more of the scene vertically compared to the 1:1 aspect ratio.

4. Widescreen (16:9) - This ratio has replaced 4:3 and is now the most common aspect ratio for TVs, monitors, and mobile phone screens (landscape). Use this aspect ratio when you want to capture more of the background (for example, scenic landscapes).

   

5. Portrait (9:16) - This ratio is widescreen but rotated. This a relatively new aspect ratio that has been popularized by short form video apps (for example, YouTube shorts). Use this for tall objects with strong vertical orientations such as buildings, trees, waterfalls, or other similar objects.

   

Photorealistic images

Different versions of the image generation model might offer a mix of artistic and photorealistic output. Use the following wording in prompts to generate more photorealistic output, based on the subject you want to generate.

Portraits

Using several keywords from the table, Imagen can generate the following portraits:

Prompt: A woman, 35mm portrait, blue and grey duotones
Model: imagen-3.0-generate-002

Prompt: A woman, 35mm portrait, film noir
Model: imagen-3.0-generate-002

Objects

Using several keywords from the table, Imagen can generate the following object images:

Prompt: leaf of a prayer plant, macro lens, 60mm
Model: imagen-3.0-generate-002

Prompt: a plate of pasta, 100mm Macro lens
Model: imagen-3.0-generate-002

Motion

Using several keywords from the table, Imagen can generate the following motion images:

Prompt: a winning touchdown, fast shutter speed, movement tracking
Model: imagen-3.0-generate-002

Prompt: A deer running in the forest, fast shutter speed, movement tracking
Model: imagen-3.0-generate-002

Wide-angle

Using several keywords from the table, Imagen can generate the following wide-angle images:

Prompt: an expansive mountain range, landscape wide angle 10mm
Model: imagen-3.0-generate-002

Prompt: a photo of the moon, astro photography, wide angle 10mm
Model: imagen-3.0-generate-002

What's next

• Check out the Veo guide to learn how to generate videos with the Gemini API.
• To learn more about Gemini 2.0 models, see Gemini models and Experimental models.