Loading video player...
In this episode of AWS Show and Tell, learn how to unlock value from your unstructured data by connecting information across content types, with a single model that delivers leading accuracy across modalities. Join us to explore Amazon Nova Multimodal Embeddings, the first unified embedding model that supports text, documents, images, video, and audio through a single model to enable cross-modal retrieval. In this session, we'll demonstrate how to efficiently convert text, documents, images, video, and audio into embeddings in a unified semantic space - eliminating the need to manage multiple specialized models. Through demos, you'll learn how to work with mixed-modality content, from documents with interleaved text and images to videos with visual, audio, and text. We'll show you implementations of reference-based image search, document retrieval, and other cross-modal applications. Learn about : Powering multimodal applications and AI agents Implementing cross-modal search and retrieval Handling mixed-modality content effectively Nova Multimodal Embeddings Announcement blog: https://aws.amazon.com/blogs/aws/amazon-nova-multimodal-embeddings-now-available-in-amazon-bedrock/ Amazon Nova Multimodal Embeddings samples: https://github.com/aws-samples/sample-demo-of-nova-mme https://github.com/aws-samples/amazon-nova-samples/tree/main/multimodal-embeddings Amazon Nova Multimodal Embeddings Model Card: https://www.amazon.science/publications/amazon-nova-multimodal-embeddings-technical-report-and-model-card More about Amazon Nova: https://aws.amazon.com/nova/ Amazon Nova model cookbook - https://github.com/aws-samples/amazon-nova-samples