Skip to main content

​
Overview

Text-to-image generation allows you to create images from natural language descriptions. Google Cloud offers two powerful options:
  • Imagen 4: Specialized text-to-image model with exceptional quality and text rendering
  • Gemini 2.5 Flash Image: Multimodal model supporting conversational image generation

​
Imagen 4 Image Generation

​
Basic Text-to-Image

Generate images using the Imagen 4 model with simple text prompts: 
from google import genai
from google.genai import types

client = genai.Client(vertexai=True, project=PROJECT_ID, location=LOCATION)

prompt = """
A white wall with two Art Deco travel posters mounted. 
First poster has the text: "NEPTUNE", tagline: "The jewel of the solar system!" 
Second poster has the text: "JUPITER", tagline: "Travel with the giants!"
"""

image = client.models.generate_images(
    model="imagen-4.0-generate-001",
    prompt=prompt,
    config=types.GenerateImagesConfig(
        aspect_ratio="16:9",
        number_of_images=1,
        image_size="2K",
        safety_filter_level="BLOCK_MEDIUM_AND_ABOVE",
        person_generation="ALLOW_ADULT",
    ),
)

# Display the generated image
display_image(image.generated_images[0].image)

​
Model Variants

Best for: High-quality images with natural lighting and photorealism
image = client.models.generate_images(
    model="imagen-4.0-generate-001",
    prompt="New York skyline at sunset",
    config=types.GenerateImagesConfig(
        number_of_images=1,
        aspect_ratio="3:4",
        image_size="2K",
    ),
)

​
Configuration Options

​
Aspect Ratios

Imagen 4 supports the following aspect ratios:
  • 1:1 - Square images
  • 3:2 / 2:3 - Standard photo dimensions
  • 4:3 / 3:4 - Classic monitor aspect ratio
  • 16:9 / 9:16 - Widescreen and mobile-friendly
  • 4:5 / 5:4 - Social media optimized
  • 21:9 - Ultra-wide cinematic

​
Image Size

Control output resolution with the image_size parameter:
  • 1K: 1024px on the longer dimension
  • 2K: 2048px on the longer dimension (higher quality)
config=types.GenerateImagesConfig(
    aspect_ratio="16:9",
    image_size="2K",  # Higher resolution
    number_of_images=1,
)

​
Number of Images

Generate multiple variations in a single request (1-4 images): 
config=types.GenerateImagesConfig(
    number_of_images=4,  # Generate 4 variations
)

​
Advanced Features

​
Multilingual Prompt Support

Imagen 4 processes prompts in multiple languages: 
prompt = """
Una pintura al óleo impresionista de una taza de café sobre una mesa 
en una cocina, con las palabras 'buenos días' escritas en una fuente 
caprichosa en la taza.
"""

image = client.models.generate_images(
    model="imagen-4.0-generate-001",
    prompt=prompt,
    config=types.GenerateImagesConfig(
        aspect_ratio="1:1",
        enhance_prompt=True,  # Enhance the prompt
    ),
)
Supported languages: English, Spanish, French, German, Portuguese, Chinese (Simplified/Traditional), Japanese, Korean, and Hindi.

​
Prompt Enhancement

Enable automatic prompt enhancement to generate more detailed descriptions: 
image = client.models.generate_images(
    model="imagen-4.0-generate-001",
    prompt="a coffee cup on a kitchen table",
    config=types.GenerateImagesConfig(
        enhance_prompt=True,  # AI will enhance your prompt
    ),
)

# View the enhanced prompt
print(image.generated_images[0].enhanced_prompt)

​
Text Rendering

Imagen 4 excels at rendering text within images: 
prompt = """
A panel of a comic strip. A cute gray cat is talking to a bulldog. 
The cat says in a talk bubble: "You really seem to enjoy going outside. 
Fascinating." Well-articulated illustration with confident lines and shading.
"""

image = client.models.generate_images(
    model="imagen-4.0-generate-001",
    prompt=prompt,
    config=types.GenerateImagesConfig(
        aspect_ratio="4:3",
        image_size="2K",
    ),
)
Use cases for text rendering:
  • Comic strips and graphic novels
  • Logos and branding materials
  • Posters and flyers
  • Infographics and flowcharts
  • Social media graphics

​
Gemini Image Generation

​
Text-to-Image

Generate images using Gemini 2.5 Flash Image: 
from google.genai.types import GenerateContentConfig, ImageConfig

MODEL_ID = "gemini-2.5-flash-image"

response = client.models.generate_content(
    model=MODEL_ID,
    contents="a cartoon infographic on flying sneakers",
    config=GenerateContentConfig(
        response_modalities=["IMAGE"],
        image_config=ImageConfig(
            aspect_ratio="9:16",
        ),
        candidate_count=1,
    ),
)

# Display the result
for part in response.candidates[0].content.parts:
    if part.inline_data:
        display(Image(data=part.inline_data.data))

​
Interleaved Text and Images

Generate tutorials or guides with mixed text and images: 
prompt = """
Create a tutorial explaining how to make a peanut butter and jelly 
sandwich in three easy steps. For each step, provide a title with 
the number of the step, an explanation, and also generate an image 
to illustrate the content.
"""

response = client.models.generate_content(
    model=MODEL_ID,
    contents=prompt,
    config=GenerateContentConfig(
        response_modalities=["TEXT", "IMAGE"],  # Both modalities
        image_config=ImageConfig(
            aspect_ratio="4:3",
        ),
    ),
)

# Display interleaved content
for part in response.candidates[0].content.parts:
    if part.text:
        display(Markdown(part.text))
    if part.inline_data:
        display(Image(data=part.inline_data.data))
This feature is perfect for creating recipes, how-to guides, product documentation, and educational content.

​
Prompt Engineering Best Practices

​
Be Specific and Descriptive

prompt = "a dog"

​
Specify Style and Mood

Include artistic style, lighting, and atmosphere: 
prompt = """
A cozy coffee shop interior, warm ambient lighting, impressionist 
painting style, soft brush strokes, muted earth tones, late afternoon 
atmosphere
"""

​
Use Technical Photography Terms

prompt = """
Portrait of a woman, 85mm lens, f/1.4 aperture, shallow depth of field,
natural window lighting, golden hour, professional studio quality
"""

​
Request Text Placement

For text rendering, be explicit about placement and style: 
prompt = """
A vintage travel poster with the text 'VISIT MARS' in bold Art Deco 
lettering at the top, and the tagline 'The Red Planet Awaits' in 
elegant script at the bottom
"""

​
Safety Controls

​
Person Generation Settings

Control whether people appear in generated images: 
config=types.GenerateImagesConfig(
    person_generation="DONT_ALLOW",     # No people
    # person_generation="ALLOW_ADULT",  # Adults only
    # person_generation="ALLOW_ALL",    # All ages
)

​
Safety Filter Levels

Adjust content filtering sensitivity: 
config=types.GenerateImagesConfig(
    safety_filter_level="BLOCK_MEDIUM_AND_ABOVE",
    # Options:
    # - BLOCK_LOW_AND_ABOVE (strictest)
    # - BLOCK_MEDIUM_AND_ABOVE (default)
    # - BLOCK_ONLY_HIGH
    # - BLOCK_NONE (permissive)
)
Choose safety settings appropriate for your application’s audience and use case.

​
Watermarking

By default, all Imagen 4 images include a SynthID digital watermark: 
config=types.GenerateImagesConfig(
    add_watermark=True,  # Default behavior
)
You can verify watermarked images using Vertex AI Studio.

​
Complete Example

Here’s a complete example generating a high-quality image: 
from google import genai
from google.genai import types
import IPython.display

# Initialize client
PROJECT_ID = "your-project-id"
LOCATION = "us-central1"
client = genai.Client(vertexai=True, project=PROJECT_ID, location=LOCATION)

# Generate image
prompt = """
Design an elegant movie poster for 'The Crimson Thread'. 
A close-up shot of two hands almost touching, with a vibrant 
crimson thread winding between their fingers. The title 'The Crimson Thread' 
should appear in flowy hand-stitched embroidery style. 
Soft-focus garden background.
"""

image = client.models.generate_images(
    model="imagen-4.0-generate-001",
    prompt=prompt,
    config=types.GenerateImagesConfig(
        aspect_ratio="1:1",
        image_size="2K",
        number_of_images=1,
        safety_filter_level="BLOCK_MEDIUM_AND_ABOVE",
        person_generation="ALLOW_ADULT",
        add_watermark=True,
    ),
)

# Display result
IPython.display.display(image.generated_images[0].image._pil_image)

​
Next Steps

Image Editing

Learn how to edit and modify existing images

Visual Q&A

Understand and analyze images with Gemini

Build docs developers (and LLMs) love

Get started for freeTalk to us