Home icon

Unified multimodal search across live events and VOD assets on AWS

Media Blog



This article describes a unified multimodal search architecture for live broadcast events and video-on-demand assets using AWS services.

  • Executes five parallel search strategies: three k-NN searches on video/audio/speech channels, one on text descriptions, and BM25 keyword matching
  • Uses TwelveLabs Marengo for cross-modal embeddings and Amazon Titan Text Embeddings for text-only vectors
  • Normalizes scores from different search legs and uses max fusion to rank results by strongest matching channel
  • Supports reverse media search with uploaded images, video clips, or audio samples matched to catalog content
  • Indexes live events within 60-90 seconds, enabling real-time search during active broadcasts
  • Spans both live (TAMS) and VOD (MAM) systems in single unified query with grouped results
  • Median search latency: ~1.1 seconds for text, ~1.2 seconds for images, ~6.5 seconds for video/audio clips

The architecture enables media professionals to discover content across multiple inventory systems using natural language without manual tagging or knowing which system holds the footage.



Go to article

The AWS News Feed is currently looking for gold sponsors. If you want to support the AWS community and reach a large audience of AWS professionals, consider sponsoring the AWS News Feed.

Related articles

Oct 27
2026
AI-enriched ingestion pipelines for searchable live and VOD media on AWS
Jun 3
2024
Implement serverless semantic search of image and live video with Amazon Titan Multimodal Embeddings
Oct 29
2026
Preparing for large-scale live streaming events on AWS — Part 1
Oct 29
2026
Preparing for large-scale live streaming events on AWS — Part 2

The AWS News Feed is currently looking for silver sponsors. If you want to support the AWS community and reach a large audience of AWS professionals, consider sponsoring the AWS News Feed.