Skip to main content
Creates an image embedding model instance that generates embeddings for image inputs using Voyage AI’s multimodal models configured for image-only processing. 
import { voyage } from 'voyage-ai-provider';

const model = voyage.imageEmbeddingModel('voyage-multimodal-3');

​
Parameters

​
modelId
VoyageMultimodalEmbeddingModelId
required
The identifier of the multimodal embedding model to use.Available models:
  • voyage-multimodal-3 - Third generation multimodal model

​
Returns

​
EmbeddingModelV3
object
An image embedding model instance that implements the AI SDK’s EmbeddingModelV3 interface.
​
modelId
string
The model identifier passed during creation
​
provider
string
The provider identifier: "voyage.image.embedding"
​
maxEmbeddingsPerCall
number
Maximum number of inputs per API call: 128
​
supportsParallelCalls
boolean
Whether parallel calls are supported: false

​
Input types

The image embedding model accepts ImageEmbeddingInput which can be:
  • Single image: string - A single image URL or base64-encoded image
  • Multiple images: string[] - Array of images combined into one embedding
  • Object format: { image: string | string[] } - Alternative format with explicit image property

​
Image formats

Images can be provided as:
  • URL: https://example.com/image.jpg (must have image extension: .jpg, .jpeg, .png, .gif, .bmp, .webp, .svg)
  • Base64: data:image/jpeg;base64,/9j/4AAQSkZJRg... (data URI with base64 encoding)

​
Usage examples

​
Generate single image embedding

Embed a single image to generate one embedding vector. 
import { voyage } from 'voyage-ai-provider';
import { embed } from 'ai';

const { embedding } = await embed({
  model: voyage.imageEmbeddingModel('voyage-multimodal-3'),
  value: 'https://i.ibb.co/nQNGqL0/beach1.jpg',
});

console.log(`Embedding length: ${embedding.length}`);

​
Generate multiple image embeddings

Embed multiple images to generate separate embedding vectors. 
import { voyage } from 'voyage-ai-provider';
import { embedMany } from 'ai';

const { embeddings } = await embedMany({
  model: voyage.imageEmbeddingModel('voyage-multimodal-3'),
  values: [
    'https://i.ibb.co/nQNGqL0/beach1.jpg',
    'https://i.ibb.co/r5w8hG8/beach2.jpg',
  ],
});

for (const [index, embedding] of embeddings.entries()) {
  console.log(`Embedding ${index}: length ${embedding.length}`);
}

​
Use base64-encoded images

You can embed images encoded as base64 data URIs. 
import { voyage } from 'voyage-ai-provider';
import { embed } from 'ai';

const getBase64Image = async (url: string) => {
  const response = await fetch(url);
  const arrayBuffer = await response.arrayBuffer();
  const base64 = Buffer.from(arrayBuffer).toString('base64');
  return `data:image/jpeg;base64,${base64}`;
};

const { embedding } = await embed({
  model: voyage.imageEmbeddingModel('voyage-multimodal-3'),
  value: await getBase64Image('https://i.ibb.co/r5w8hG8/beach2.jpg'),
});

​
Use object format

You can use the object format with explicit image property. 
import { voyage } from 'voyage-ai-provider';
import { embedMany } from 'ai';

const { embeddings } = await embedMany({
  model: voyage.imageEmbeddingModel('voyage-multimodal-3'),
  values: [
    {
      image: 'https://i.ibb.co/nQNGqL0/beach1.jpg',
    },
    {
      image: await getBase64Image('https://i.ibb.co/r5w8hG8/beach2.jpg'),
    },
  ],
});

​
Combine multiple images into one embedding

You can combine multiple images into a single embedding by passing an array. This is useful for representing related images as a single vector. 
import { voyage } from 'voyage-ai-provider';
import { embedMany } from 'ai';

const { embeddings } = await embedMany({
  model: voyage.imageEmbeddingModel('voyage-multimodal-3'),
  values: [
    {
      image: [
        'https://i.ibb.co/nQNGqL0/beach1.jpg',
        'https://i.ibb.co/r5w8hG8/beach2.jpg',
      ],
    },
  ],
});

console.log(`Generated ${embeddings.length} embedding(s)`);

​
Multiple embeddings with multiple images

Generate multiple embeddings, each combining multiple images. 
import { voyage } from 'voyage-ai-provider';
import { embedMany } from 'ai';

const { embeddings } = await embedMany({
  model: voyage.imageEmbeddingModel('voyage-multimodal-3'),
  values: [
    {
      image: [
        'https://i.ibb.co/nQNGqL0/beach1.jpg',
        'https://i.ibb.co/r5w8hG8/beach2.jpg',
      ],
    },
    {
      image: [
        'https://i0.wp.com/blog.voyageai.com/wp-content/uploads/2024/11/Slide-1.png',
      ],
    },
  ],
});

console.log(`Generated ${embeddings.length} embeddings`);

​
Use provider options

You can customize the embedding behavior using provider-specific options. 
import { voyage } from 'voyage-ai-provider';
import { embed } from 'ai';
import type { VoyageMultimodalEmbeddingOptions } from 'voyage-ai-provider';

const { embedding } = await embed({
  model: voyage.imageEmbeddingModel('voyage-multimodal-3'),
  value: 'https://i.ibb.co/nQNGqL0/beach1.jpg',
  providerOptions: {
    voyage: {
      inputType: 'document',
      truncation: true,
    } satisfies VoyageMultimodalEmbeddingOptions,
  },
});

​
Provider options

You can pass Voyage-specific options through the providerOptions parameter:
​
providerOptions.voyage.inputType
'query' | 'document'
The input type for the embeddings. Defaults to "query".When specified, Voyage automatically prepends a prompt to your inputs before vectorizing them, creating vectors more tailored for retrieval/search tasks.
  • query: Prepends “Represent the query for retrieving supporting documents: ”
  • document: Prepends “Represent the document for retrieval: ”
For retrieval/search purposes where a query is used to search through documents, we recommend specifying whether your inputs are queries or documents.
​
providerOptions.voyage.outputEncoding
'base64'
The data type for the resulting output embeddings.If not specified (defaults to null), the embeddings are represented as a list of floating-point numbers.If 'base64', the embeddings are represented as a Base64-encoded NumPy array of single-precision floats.
​
providerOptions.voyage.truncation
boolean
Whether to truncate the input to fit within the context length. Defaults to true.

Build docs developers (and LLMs) love

Get started for freeTalk to us