Text-to-Image

Generate images from text prompts using Nano Banana and Imagen 4 models, with support forsearch grounding, text rendering, reference images, and configurable safety controls.

Available Models

Model ID	Name	Engine	Speed	Max Resolution	Max Output Images	Key Capability
vertex/nano-banana-flash	Nano Banana Flash	Gemini Flash	Fast	4K	10	Search grounding
vertex/nano-banana-pro	Nano Banana Pro	Gemini Pro	Standard	4K	10	Text rendering, character consistency
vertex/imagen-4	Imagen 4	Imagen 4.0	Standard	2K	4	Balanced quality and adherence
vertex/imagen-4-fast	Imagen 4 Fast	Imagen 4.0 Fast	Fastest	1K	4	Lowest cost, highest throughput
vertex/imagen-4-ultra	Imagen 4 Ultra	Imagen 4.0 Ultra	Slowest	2K	4	Highest quality, complex prompts

Nano Banana Flash

Model ID: vertex/nano-banana-flash | Engine: Gemini Flash

Optimized for fast generation, search-grounded accuracy, and high-volume workflows.

Minimal Request

{
  "model": "vertex/nano-banana-flash",
  "prompt": "A cozy Japanese ramen shop on a rainy night, steam rising from bowls, warm ambient lighting"
}

With Parameters

{
  "model": "vertex/nano-banana-flash",
  "prompt": "A hyperrealistic macro photograph of a butterfly on a lavender flower, water droplets on wings, golden hour light",
  "aspect_ratio": "4:3",
  "quality": "2K",
  "num_images": 1,
  "temperature": 1.0,
  "seed": 42
}

Search Grounding

Set use_search_grounding: true to incorporate real-time Google Search context. This improves factual and visual accuracy for real-world subjects such as architecture, landmarks, and products.

{
  "model": "vertex/nano-banana-flash",
  "prompt": "The interior of the Pantheon in Rome with sunlight streaming through the oculus, accurate architecture and proportions",
  "aspect_ratio": "4:3",
  "quality": "2K",
  "use_search_grounding": true
}

Batch Variations

Generate multiple variations of a single prompt by setting num_images (up to 10). Increase temperature above 1.0 for greater variation between outputs.

{
  "model": "vertex/nano-banana-flash",
  "prompt": "A minimalist flat illustration of a coffee cup with steam, pastel color palette, icon style",
  "aspect_ratio": "1:1",
  "quality": "1K",
  "num_images": 4,
  "temperature": 1.1
}

Nano Banana Pro

Model ID: vertex/nano-banana-pro | Engine: Gemini Pro

Optimized for production-quality output, multi-language text rendering in images, 4K resolution, and consistent character or product depiction across multiple requests.

Minimal Request

{
  "model": "vertex/nano-banana-pro",
  "prompt": "A luxury watch on a black marble surface, dramatic spotlight lighting, product photography"
}

With Parameters

{
  "model": "vertex/nano-banana-pro",
  "prompt": "A cinema-quality portrait of a woman in a traditional Japanese kimono standing in a cherry blossom garden, shallow depth of field, golden hour lighting",
  "aspect_ratio": "3:4",
  "quality": "4K",
  "num_images": 1,
  "temperature": 0.9,
  "seed": 12345
}

Text Rendering

Nano Banana Pro renders text within generated images with high accuracy across multiple languages.

{
  "model": "vertex/nano-banana-pro",
  "prompt": "A clean coffee shop menu board with the text 'Espresso $3.50', 'Latte $4.50', 'Cappuccino $4.00' written in elegant chalk lettering on a dark background",
  "aspect_ratio": "3:4",
  "quality": "2K",
  "temperature": 0.8
}

Reference Images

Pass reference images for character or product consistency across multiple outputs. Each reference includes a url and a description guiding the model on what to match.

{
  "model": "vertex/nano-banana-pro",
  "prompt": "Generate this character standing in a futuristic neon-lit city plaza at night. Keep exact face, hairstyle, and outfit from the references.",
  "reference_images": [
    {
      "url": "https://your-cdn.com/character-front.jpg",
      "description": "Character front view — match face, hair color, and facial features exactly"
    },
    {
      "url": "https://your-cdn.com/character-outfit.jpg",
      "description": "Character outfit — match the denim jacket, white t-shirt, and black jeans"
    }
  ],
  "quality": "2K",
  "aspect_ratio": "9:16",
  "temperature": 0.85
}

Imagen 4

Model ID: vertex/imagen-4 | Engine: Imagen 4.0

General-purpose generation with high prompt adherence, multilingual support, and configurable safety filtering and person-generation policies.

Minimal Request

{
  "model": "vertex/imagen-4",
  "prompt": "A hyperrealistic close-up of a dew-covered spider web at dawn, macro photography"
}

With Parameters

{
  "model": "vertex/imagen-4",
  "prompt": "A sweeping landscape of the Scottish Highlands at golden hour, rolling heather-covered hills, dramatic storm clouds breaking, a lone castle in the distance",
  "aspect_ratio": "16:9",
  "num_images": 1,
  "image_size": "2K"
}

Safety Controls

Imagen models support fine-grained content policies via person_generation and safety_filter_level.

{
  "model": "vertex/imagen-4",
  "prompt": "A professional portrait of a businessperson in a modern office, natural window light, confident expression",
  "aspect_ratio": "3:4",
  "image_size": "2K",
  "num_images": 2,
  "output_compression_quality": 90,
  "person_generation": "allow_adult",
  "safety_filter_level": "block_medium_and_above",
  "enhance_prompt": false,
  "add_watermark": false
}

Imagen 4 Fast

Model ID: vertex/imagen-4-fast | Engine: Imagen 4.0 Fast

Optimized for bulk generation, rapid prototyping, and cost-sensitive workloads. Output is fixed at approximately 1K resolution.

Minimal Request

{
  "model": "vertex/imagen-4-fast",
  "prompt": "A vibrant flat illustration of a bicycle in a sunny park, digital art style"
}

With Parameters

{
  "model": "vertex/imagen-4-fast",
  "prompt": "A simple icon of a shield with a checkmark, flat design, blue and white, clean vector style",
  "aspect_ratio": "1:1",
  "num_images": 4
}

Async Mode

Set mode: "async" to return a task_id immediately instead of waiting for generation to complete. Poll GET /tasks/{task_id} for results.

{
  "model": "vertex/imagen-4-fast",
  "prompt": "A colorful abstract geometric background pattern, modern gradient, suitable for mobile app splash screen",
  "aspect_ratio": "9:16",
  "num_images": 4,
  "mode": "async"
}

Imagen 4 Ultra

Model ID: vertex/imagen-4-ultra | Engine: Imagen 4.0 Ultra

Highest quality output with high prompt adherence for complex multi-element compositions. Slower and higher cost than other Imagen 4 variants.

Minimal Request

{
  "model": "vertex/imagen-4-ultra",
  "prompt": "A detailed oil painting of a medieval port city at twilight, torchlit cobblestone streets, galleons in harbor, figures in period clothing"
}

With Parameters

{
  "model": "vertex/imagen-4-ultra",
  "prompt": "A hyperrealistic underwater scene: a coral reef ecosystem at depth, shafts of turquoise sunlight filtering through the surface, diverse marine life including clownfish, sea turtles, and manta rays",
  "aspect_ratio": "16:9",
  "image_size": "2K",
  "num_images": 1
}

Maximum Quality

{
  "model": "vertex/imagen-4-ultra",
  "prompt": "A sweeping aerial panorama of the Dolomites in Italy at sunrise, snow-capped peaks glowing pink and orange, deep valleys in morning shadow, a winding mountain road visible far below",
  "aspect_ratio": "16:9",
  "image_size": "2K",
  "num_images": 1,
  "enhance_prompt": false,
  "add_watermark": false,
  "safety_filter_level": "block_medium_and_above"
}

Supported Aspect Ratios

Ratio	Nano Banana Flash	Nano Banana Pro	Imagen 4 / Fast / Ultra
1:1	Yes	Yes	Yes
4:3	Yes	Yes	Yes
3:4	Yes	Yes	Yes
16:9	Yes	Yes	Yes
9:16	Yes	Yes	Yes
3:2	Yes	Yes	—
2:3	Yes	Yes	—
21:9	Yes	Yes	—
1:4	Yes	Yes	—
4:1	Yes	Yes	—
5:4	Yes	Yes	—
4:5	Yes	Yes	—