or run

npx @tessl/cli init

Version

Tile

Overview

Evals

Files

docs

built-in-tools.md image-generation.md index.md language-models.md provider-setup.md text-embeddings.md

tile.json

Image Generation

Image generation capabilities using Google's Imagen models for creating high-quality images from text prompts with configurable settings and output options.

Capabilities

Image Model Access

Get image generation model instances for creating images from text descriptions.

/**
 * Get an image generation model instance
 * @param modelId - Google image model identifier
 * @param settings - Optional image generation settings
 * @returns ImageModelV2 instance
 */
image(modelId: GoogleGenerativeAIImageModelId, settings?: GoogleGenerativeAIImageSettings): ImageModelV2<
  GoogleGenerativeAIImageProviderOptions
>;

/**
 * Get an image generation model instance (alias for image)
 * @param modelId - Google image model identifier  
 * @param settings - Optional image generation settings
 * @returns ImageModelV2 instance
 */
imageModel(modelId: GoogleGenerativeAIImageModelId, settings?: GoogleGenerativeAIImageSettings): ImageModelV2<
  GoogleGenerativeAIImageProviderOptions
>;

Usage Examples:

import { google } from "@ai-sdk/google";
import { generateImage } from "ai";

// Basic image generation
const result = await generateImage({
  model: google.image("imagen-3.0-generate-002"),
  prompt: "A beautiful sunset over mountains with a lake in the foreground",
});

console.log("Generated image URL:", result.image.url);

// Using imageModel method (alias for image)
const imageModel = google.imageModel("imagen-3.0-generate-002");
const result2 = await generateImage({
  model: imageModel,
  prompt: "A futuristic city with flying cars",
});

// Multiple images generation
const result = await generateImage({
  model: google.image("imagen-3.0-generate-002"),
  prompt: "A cute robot playing with a ball",
  n: 3, // Generate 3 images
});

result.images.forEach((image, index) => {
  console.log(`Image ${index + 1}:`, image.url);
});

Image Model IDs

Supported Google image generation model identifiers.

type GoogleGenerativeAIImageModelId =
  | "imagen-3.0-generate-002"
  | (string & {});

Model Information:

imagen-3.0-generate-002: Google's latest Imagen model with high-quality image generation capabilities

Image Provider Options

Configuration options for image generation behavior and output settings.

interface GoogleGenerativeAIImageProviderOptions {
  /** Custom settings for the image model */
  maxImagesPerCall?: number;
}

interface GoogleGenerativeAIImageSettings {
  /** Override the maximum number of images per call (default: 4) */
  maxImagesPerCall?: number;
}

Usage Examples:

import { google } from "@ai-sdk/google";
import { generateImage } from "ai";

// Custom image generation settings
const imageModel = google.image("imagen-3.0-generate-002");

const result = await generateImage({
  model: imageModel,
  prompt: "A fantasy landscape with magical creatures",
  n: 2,
  providerOptions: {
    maxImagesPerCall: 2, // Override default maximum
  },
});

// Access generated images
for (const image of result.images) {
  console.log("Image URL:", image.url);
  console.log("Alt text:", image.alt);
}

Image Generation Features

Single Image Generation

Generate a single image from a text prompt.

import { google } from "@ai-sdk/google";
import { generateImage } from "ai";

const result = await generateImage({
  model: google.image("imagen-3.0-generate-002"),
  prompt: "A serene Japanese garden with cherry blossoms and a koi pond",
});

// Access the generated image
console.log("Image URL:", result.image.url);
console.log("Alt description:", result.image.alt);

Multiple Image Generation

Generate multiple variations of images from the same prompt.

import { google } from "@ai-sdk/google";
import { generateImage } from "ai";

const result = await generateImage({
  model: google.image("imagen-3.0-generate-002"),
  prompt: "A modern office space with natural lighting",
  n: 4, // Generate 4 variations
});

// Process all generated images
result.images.forEach((image, index) => {
  console.log(`Variation ${index + 1}:`);
  console.log("  URL:", image.url);
  console.log("  Description:", image.alt);
});

console.log(`Generated ${result.images.length} images total`);

Advanced Prompting

Create detailed prompts for specific image characteristics.

import { google } from "@ai-sdk/google";
import { generateImage } from "ai";

// Detailed artistic prompt
const artisticResult = await generateImage({
  model: google.image("imagen-3.0-generate-002"),
  prompt: `A photorealistic portrait of a wise old wizard with a long white beard, 
           wearing deep blue robes with golden embroidery, holding a glowing crystal staff, 
           standing in an ancient library filled with floating books and magical artifacts, 
           dramatic lighting, high detail, fantasy art style`,
});

// Architectural prompt
const architecturalResult = await generateImage({
  model: google.image("imagen-3.0-generate-002"),
  prompt: `Modern minimalist house design with floor-to-ceiling windows, 
           clean geometric lines, surrounded by lush greenery, 
           natural materials like wood and stone, bright daylight, 
           architectural photography style`,
});

// Abstract art prompt
const abstractResult = await generateImage({
  model: google.image("imagen-3.0-generate-002"),
  prompt: `Abstract digital art with flowing organic shapes in vibrant colors, 
           gradients of blue and purple merging into golden highlights, 
           dreamlike and ethereal composition, high resolution`,
});

Configuration Options

Maximum Images Per Call

Control the maximum number of images that can be generated in a single call.

import { google } from "@ai-sdk/google";
import { generateImage } from "ai";

// Custom provider with specific image settings
const customGoogle = createGoogleGenerativeAI({
  apiKey: process.env.GOOGLE_GENERATIVE_AI_API_KEY,
});

const result = await generateImage({
  model: customGoogle.image("imagen-3.0-generate-002"),
  prompt: "A collection of geometric patterns",
  n: 3,
  providerOptions: {
    maxImagesPerCall: 3, // Ensure we can generate 3 images
  },
});

Error Handling

Handle image generation errors and limitations.

import { google } from "@ai-sdk/google";
import { generateImage } from "ai";

try {
  const result = await generateImage({
    model: google.image("imagen-3.0-generate-002"),
    prompt: "A beautiful landscape painting",
    n: 5, // This might exceed maxImagesPerCall
  });
  
  console.log(`Successfully generated ${result.images.length} images`);
} catch (error) {
  if (error.code === "INVALID_ARGUMENT") {
    console.error("Invalid request parameters:", error.message);
  } else if (error.code === "QUOTA_EXCEEDED") {
    console.error("API quota exceeded:", error.message);
  } else {
    console.error("Image generation failed:", error.message);
  }
}

Best Practices

Prompt Engineering

Be Specific: Include details about style, composition, lighting, and atmosphere
Use Descriptive Adjectives: Add adjectives that describe quality (photorealistic, detailed, vibrant)
Specify Art Styles: Mention specific art styles or photography techniques when desired
Include Technical Details: Specify resolution, lighting conditions, or camera angles for realistic images

Performance Optimization

Batch Requests: Generate multiple variations in a single call when possible
Appropriate Image Count: Use the default maxImagesPerCall (4) unless you need fewer images
Prompt Length: Keep prompts descriptive but concise for optimal results

Content Guidelines

Follow Google's content policies for image generation
Avoid generating images of real people without consent
Be mindful of copyrighted characters or brands
Consider cultural sensitivity in image descriptions

Example: Complete Image Generation Workflow

import { google } from "@ai-sdk/google";
import { generateImage } from "ai";
import fs from "fs/promises";

async function generateAndSaveImages() {
  try {
    // Generate multiple image variations
    const result = await generateImage({
      model: google.image("imagen-3.0-generate-002"),
      prompt: `A cozy reading nook by a window with warm sunlight, 
               comfortable armchair, stacks of books, steaming coffee cup, 
               indoor plants, soft textures, inviting atmosphere`,
      n: 3,
    });

    // Process and save each image
    for (let i = 0; i < result.images.length; i++) {
      const image = result.images[i];
      console.log(`Processing image ${i + 1}:`);
      console.log(`  URL: ${image.url}`);
      console.log(`  Description: ${image.alt}`);
      
      // Optional: Download and save the image
      // const response = await fetch(image.url);
      // const buffer = await response.arrayBuffer();
      // await fs.writeFile(`reading-nook-${i + 1}.png`, Buffer.from(buffer));
    }

    return result;
  } catch (error) {
    console.error("Failed to generate images:", error);
    throw error;
  }
}