generate_images

Generate images using Imagen models

Method Signature

def generate_images(
    self,
    *,
    model: str,
    prompt: str,
    config: Optional[GenerateImagesConfig] = None,
) -> GenerateImagesResponse

async def generate_images(
    self,
    *,
    model: str,
    prompt: str,
    config: Optional[GenerateImagesConfig] = None,
) -> GenerateImagesResponse

Description

Generates images based on a text description using Imagen models. Supports various image generation parameters including aspect ratio, safety filters, and output format control.

Parameters

model · str · required

The Imagen model to use for generation.

Examples:

• 'imagen-3.0-generate-002'
• 'imagen-3.0-fast-generate-001'
• 'imagen-3.0-generate-001'

prompt · str · required

A text description of the images to generate.

Example: 'A serene mountain landscape at sunset with a lake in the foreground'

config · GenerateImagesConfig

Configuration for image generation.

Show config properties

number_of_images · int

Number of images to generate (default: 1, max: 4)

aspect_ratio · str

Aspect ratio of generated images.

Options:

• '1:1' - Square (default)
• '16:9' - Landscape
• '9:16' - Portrait
• '4:3'
• '3:4'

negative_prompt · str

Text describing what to avoid in the image.

Example: 'blurry, low quality, distorted'

Vertex AI only

seed · int

Random seed for reproducible generation.

Vertex AI only

guidance_scale · float

How closely to follow the prompt (1.0 to 20.0). Higher values = more faithful to prompt.

safety_filter_level · str

Safety filter strictness level.

Options:

• 'BLOCK_LOW_AND_ABOVE'
• 'BLOCK_MEDIUM_AND_ABOVE'
• 'BLOCK_ONLY_HIGH'
• 'BLOCK_NONE' (Vertex AI only)

person_generation · str

Policy for generating images of people.

Options:

• 'DONT_ALLOW' - Don’t generate people
• 'ALLOW_ADULT' - Allow adult depictions only
• 'ALLOW_ALL' (Vertex AI only)

include_safety_attributes · bool

Include safety classification in response (default: false)

include_rai_reason · bool

Include RAI (Responsible AI) filtering reasons (default: false)

language · str

Language code for prompt interpretation (e.g., 'en', 'es', 'ja')

output_mime_type · str

Output image format.

Options:

• 'image/jpeg'
• 'image/png' (default)

output_compression_quality · int

JPEG compression quality (1-100). Only applies when output_mime_type is 'image/jpeg'.

add_watermark · bool

Add a watermark to generated images (default: true)

Vertex AI only

image_size · str

Image dimensions.

Examples: '1024', '512'

enhance_prompt · bool

Automatically enhance the prompt for better results

Vertex AI only

output_gcs_uri · str

Google Cloud Storage URI to save generated images

Example: 'gs://my-bucket/images/'

Vertex AI only

labels · dict[str, str]

Labels to attach to the generation request

Vertex AI only

Response

generated_images · list[GeneratedImage]

List of generated images.

Show GeneratedImage properties

image · Image

The generated image object with methods:

• .show() - Display the image
• .save(path) - Save to file
• .to_pil() - Convert to PIL Image
• .uri - GCS URI if saved to cloud storage
• .image_bytes - Raw image bytes

rai_filtered_reason · str

Reason if image was filtered by Responsible AI filters

safety_attributes · SafetyAttributes

Safety classifications (if include_safety_attributes=True)

Show properties

scores · list[float]

Safety scores for different categories

blocked · bool

Whether image was blocked

enhanced_prompt · str

The enhanced prompt used (if enhance_prompt=True)

Vertex AI only

positive_prompt_safety_attributes · SafetyAttributes

Safety attributes for the prompt itself

Code Examples

Basic Image Generation

from google import genai

client = genai.Client(vertexai=True, project='my-project', location='us-central1')

response = client.models.generate_images(
    model='imagen-3.0-generate-002',
    prompt='A majestic mountain landscape at sunset'
)

# Display the image
response.generated_images[0].image.show()

# Save the image
response.generated_images[0].image.save('mountain.png')

Multiple Images with Custom Settings

from google.genai import types

response = client.models.generate_images(
    model='imagen-3.0-generate-002',
    prompt='A cute robot playing with a cat',
    config=types.GenerateImagesConfig(
        number_of_images=4,
        aspect_ratio='16:9',
        guidance_scale=15.0,
        include_rai_reason=True,
    )
)

# Process all generated images
for i, gen_image in enumerate(response.generated_images):
    gen_image.image.save(f'robot_cat_{i}.png')
    if gen_image.rai_filtered_reason:
        print(f"Image {i} filtered: {gen_image.rai_filtered_reason}")

With Negative Prompt and Safety Settings

response = client.models.generate_images(
    model='imagen-3.0-generate-002',
    prompt='A photorealistic portrait of a person smiling',
    config=types.GenerateImagesConfig(
        negative_prompt='blurry, distorted, low quality, cartoon',
        person_generation='ALLOW_ADULT',
        safety_filter_level='BLOCK_MEDIUM_AND_ABOVE',
        include_safety_attributes=True,
    )
)

if response.generated_images:
    print(f"Safety attributes: {response.generated_images[0].safety_attributes}")
    response.generated_images[0].image.show()

Save to Cloud Storage

response = client.models.generate_images(
    model='imagen-3.0-generate-002',
    prompt='An abstract geometric pattern',
    config=types.GenerateImagesConfig(
        number_of_images=2,
        output_gcs_uri='gs://my-bucket/images/',
        output_mime_type='image/jpeg',
        output_compression_quality=90,
    )
)

for gen_image in response.generated_images:
    print(f"Image saved to: {gen_image.image.uri}")

High-Quality Output Settings

response = client.models.generate_images(
    model='imagen-3.0-generate-002',
    prompt='A detailed illustration of a steampunk city',
    config=types.GenerateImagesConfig(
        aspect_ratio='16:9',
        guidance_scale=18.0,
        image_size='1024',
        output_mime_type='image/png',
        enhance_prompt=True,
        seed=42,  # For reproducibility
    )
)

print(f"Enhanced prompt: {response.generated_images[0].enhanced_prompt}")
response.generated_images[0].image.show()

Async Usage

import asyncio
from google import genai

client = genai.Client(vertexai=True, project='my-project', location='us-central1')

async def generate():
    response = await client.aio.models.generate_images(
        model='imagen-3.0-generate-002',
        prompt='A futuristic cityscape at night',
        config={'number_of_images': 2}
    )
    
    for i, gen_image in enumerate(response.generated_images):
        gen_image.image.save(f'city_{i}.png')

asyncio.run(generate())

Notes

• Generation typically takes 5-30 seconds depending on the model and settings
• Some configuration options are only available on Vertex AI
• Images may be filtered by safety systems if they violate content policies
• Use include_rai_reason=True to understand why images were filtered
• The enhance_prompt feature can significantly improve results but may alter your intent
• Higher guidance_scale values make the model follow the prompt more closely but may reduce creativity

Related Methods

See the Imagen guide for more information on editing and upscaling images with edit_image and upscale_image methods.