How Condé Nast built multimodal video discovery with Amazon Bedrock

This article describes how Condé Nast partnered with AWS to build an AI-powered multimodal video discovery solution that dramatically accelerates content discovery across their 140,000+ video library.

• Reduced discovery time from 250 minutes to under 2 minutes per task using semantic search across transcripts, visuals, and audio
• Built on Amazon Bedrock with TwelveLabs Marengo embedding model for multimodal understanding
• Decoupled ingestion and serving planes enable independent scaling and evolution
• Supports intent-based search, image-based queries, and precise timestamp results
• Achieved 99.2% reduction in discovery time and estimated $800,000 annual operational savings
• Multi-AZ architecture ensures high availability for editorial teams throughout the workday

The solution demonstrates how semantic search and multimodal embeddings can transform content discovery workflows for media organizations managing large video libraries.