AmazonTitan
Titan Multimodal Embeddings G1
Titan Multimodal Embeddings G1 puts text and images into a shared semantic space. The Bedrock card lists Active status.
What the release provides
The API model is amazon.titan-embed-image-v1. It supports retrieval and recommendation uses with embedding output.Titan Multimodal Embeddings G1 Bedrock model card
How to use it well
Practical guidance. For a product-image catalog, compare visual similarity with text descriptions and known item attributes. Similar shape or color can retrieve the wrong product. Keep exact identifiers outside the vector search and validate them before acting on a match.
Text and image limits
The guide specifies English text input up to 256 tokens and images up to 25 MB and 2048 by 2048 pixels. Vector dimensions can be 1,024, 384 or 256.Titan multimodal embedding specifications
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | API |
| Model identifiers | amazon.titan-embed-image-v1 |