ContextQMD
Back to Promptfoo

google-imagen (Google Imagen)

examples/google-imagen/README.md

0.121.94.3 KB
Original Source

google-imagen (Google Imagen)

This example demonstrates Google image generation models, including both Imagen and Gemini native image generation.

You can run this example with:

bash
npx promptfoo@latest init --example google-imagen
cd google-imagen

Prerequisites

You can use Imagen models through either Google AI Studio or Vertex AI:

Option 1: Google AI Studio (Quick Start, Limited Features)

  • Get an API key from Google AI Studio
  • Supports: Imagen 4 models only
  • Limitations: No seed or addWatermark parameters

Option 2: Vertex AI (Full Features)

  • Google Cloud project with billing enabled
  • Vertex AI API enabled
  • Authentication via gcloud auth application-default login
  • Supports: All Imagen models and all parameters

Setup

For Google AI Studio:

bash
# Set your API key (Unix/Linux/macOS)
export GOOGLE_API_KEY=your-api-key

# For Windows Command Prompt:
# set GOOGLE_API_KEY=your-api-key

# For Windows PowerShell:
# $env:GOOGLE_API_KEY="your-api-key"

For Vertex AI:

bash
# Enable Vertex AI API
gcloud services enable aiplatform.googleapis.com

# Authenticate
gcloud auth application-default login

# Set project ID (Unix/Linux/macOS)
export GOOGLE_PROJECT_ID=your-project-id

# For Windows Command Prompt:
# set GOOGLE_PROJECT_ID=your-project-id

# For Windows PowerShell:
# $env:GOOGLE_PROJECT_ID="your-project-id"

Environment Variables

  • GOOGLE_API_KEY - Google AI Studio API key (Option 1)
  • GOOGLE_PROJECT_ID - Your Google Cloud project ID (Option 2)

Available Models

Imagen Models (use google:image: prefix)

  • imagen-4.0-ultra-generate-preview-06-06 - Ultra quality ($0.06/image)
  • imagen-4.0-generate-preview-06-06 - Standard quality ($0.04/image)
  • imagen-4.0-fast-generate-preview-06-06 - Fast generation ($0.02/image)
  • imagen-3.0-generate-002 - Imagen 3.0 ($0.04/image)

Gemini Native Image Generation

  • google:gemini-3.1-flash-image-preview - Gemini 3.1 Flash (Nano Banana 2) with native image generation (~$0.067/image at 1K)
  • google:gemini-3-pro-image-preview - Gemini 3 Pro with native image generation (~$0.05/image, estimated)
  • google:gemini-2.5-flash-image - Gemini 2.5 Flash with image generation (~$0.04/image)

Running the Example

bash
npx promptfoo@latest eval

Notes

  • Imagen models are available through both Google AI Studio and Vertex AI
  • Google AI Studio:
    • Uses API key authentication (quick start)
    • Only supports Imagen 4 models (4.0 preview models)
    • Limited parameter support:
      • No seed or addWatermark parameters
      • Only block_low_and_above safety filter level
  • Vertex AI:
    • Requires authentication via gcloud auth application-default login
    • Supports all Imagen models (both 3.0 and 4.0)
    • Full parameter support:
      • All safety filter levels (block_most, block_some, block_few, block_fewest)
      • seed for deterministic generation
      • addWatermark control
    • You must provide a Google Cloud project ID either via:
      • GOOGLE_PROJECT_ID environment variable
      • projectId in the provider config
  • Seed and watermark parameters are mutually exclusive (Vertex AI only)

Advanced Configuration

See promptfooconfig-advanced.yaml for examples of platform-specific configurations that take advantage of each platform's unique capabilities.

Gemini Native Image Generation

Gemini models can generate images natively using the generateContent API with responseModalities set to include images. This is different from Imagen which uses a dedicated image generation endpoint.

See promptfooconfig-gemini.yaml for Gemini native image generation examples. See promptfooconfig-gemini-grounding.yaml for Google Search-grounded Gemini image generation.

Key differences from Imagen:

  • Uses the same google: provider namespace as Gemini chat (models with -image in the name automatically enable image generation)
  • Supports additional aspect ratios: 1:4, 1:8, 2:3, 3:2, 4:1, 4:5, 5:4, 8:1, 21:9
  • Supports image resolution: 512px (Gemini 3.1), 1K, 2K, 4K
  • Can return both text and images in the same response
  • Uses the same authentication as Gemini chat models
  • Supports Google Search grounding via tools: [{ googleSearch: {} }]