Skip to main content

​
Introduction to Vision AI

Google Cloud offers powerful generative AI capabilities for creating and editing images through two primary model families:

Imagen

Google’s state-of-the-art text-to-image model for high-quality image generation

Gemini Image Generation

Multimodal conversational image generation and editing with Gemini models

​
Imagen on Vertex AI

Imagen brings Google’s state-of-the-art generative AI capabilities to application developers. The Imagen family includes:
  • Imagen 4: Google’s highest quality text-to-image model with exceptional detail, improved prompt adherence, and advanced text rendering
  • Imagen 4 Fast: Optimized for lower latency with brighter images and higher contrast
  • Imagen 4 Ultra: Exceptional quality with enhanced photorealism
  • Imagen 3: Previous generation with robust editing and customization capabilities

​
Key Capabilities

Text-to-Image Generation

Generate high-quality images from natural language descriptions

Image Editing

Modify existing images with inpainting, outpainting, and mask-based editing

Text Rendering

Accurately render text within images for posters, comics, and logos

Multilingual Support

Process prompts in 10+ languages including English, Spanish, Japanese, and more

​
Gemini Image Generation

Gemini 2.5 Flash Image (also known as “Nano Banana”) is a powerful multimodal model that combines conversational AI with image generation and editing capabilities.

​
Unique Features

  • Conversational Image Editing: Iteratively refine images through natural conversation
  • Interleaved Content: Generate sequences mixing text and images (e.g., step-by-step tutorials)
  • Multi-Reference Editing: Combine multiple input images to create new compositions
  • Subject Customization: Apply specific subjects or styles from reference images
Gemini 2.5 Flash Image is ideal for applications requiring iterative design workflows, interactive image creation, or tightly integrated text and visual content.

​
Getting Started

1

Set up your environment

Install the Google Gen AI SDK for Python:
pip install --upgrade google-genai
2

Configure authentication

Set up your Google Cloud project and enable the Vertex AI API:
from google import genai
from google.genai import types

PROJECT_ID = "your-project-id"
LOCATION = "us-central1"

client = genai.Client(vertexai=True, project=PROJECT_ID, location=LOCATION)
3

Generate your first image

Start creating images with a simple text prompt:
# Using Imagen 4
image = client.models.generate_images(
    model="imagen-4.0-generate-001",
    prompt="a serene mountain landscape at sunset",
    config=types.GenerateImagesConfig(
        aspect_ratio="16:9",
        number_of_images=1,
    ),
)

​
Model Selection Guide

ModelBest ForLatencyQuality
Imagen 4High-quality images with text renderingMediumHighest
Imagen 4 FastQuick iterations, prototypingLowHigh
Imagen 4 UltraProfessional-grade photorealismHighExceptional
Gemini 2.5 Flash ImageConversational editing, tutorialsMediumHigh

​
Safety and Watermarking

All images generated with these models include:
  • SynthID Watermark: Digital watermark for identifying AI-generated content
  • Safety Filters: Configurable content filtering to block inappropriate content
  • Person Generation Controls: Fine-grained control over generating images with people
Always configure appropriate safety filters for your application’s use case. Available levels include BLOCK_LOW_AND_ABOVE, BLOCK_MEDIUM_AND_ABOVE, BLOCK_ONLY_HIGH, and BLOCK_NONE.

​
Learn More

Image Generation

Text-to-image generation techniques and prompt engineering

Image Editing

Inpainting, outpainting, and advanced editing workflows

Visual Q&A

Image understanding and visual question answering

​
Supported Locations

Imagen and Gemini Image models are available in multiple regions. For the most up-to-date list of supported locations, see the Vertex AI locations documentation.

Build docs developers (and LLMs) love

Get started for freeTalk to us