{
  "id": 10727214,
  "title": "How Condé Nast built multimodal video discovery with Amazon Bedrock",
  "url": "https://urgent.news/2026/09/29/how-conde-nast-built-multimodal-video-discovery-with-amazon-bedrock",
  "topic": "ai",
  "section": "AI",
  "published": "2026-09-29T15:55:17.000Z",
  "source": {
    "name": "AWS Machine Learning",
    "slug": "aws-machine-learning",
    "url": "https://aws.amazon.com/blogs/machine-learning/how-conde-nast-built-multimodal-video-discovery-with-amazon-bedrock/"
  },
  "original_language": "en",
  "account": "Condé Nast, a media company with brands like Vogue and GQ, faced a challenge with their video content discovery process. They were spending an average of 250 minutes per task to manually search through over 140,000 videos, relying on titles and descriptions which often didn't yield relevant clips. This slow process was causing operational drag and making it difficult to discover underutilized content.\n\nTo address this issue, Condé Nast partnered with the AWS Generative AI Innovation Center (GenAIIC) to build an AI-powered multimodal video discovery solution. This solution was built on Amazon Bedrock and Amazon OpenSearch Service, and it utilizes intent-based semantic search across video transcripts, visual elements, and audio.\n\nThe team chose the TwelveLabs Marengo embedding model for its ability to jointly encode visual, audio, and transcript signals. This model powers all five capabilities of the solution, which include intent-based search, multimodal understanding, image-based queries, typo tolerance, and timestamp precision.\n\nThe architecture of the solution is decoupled into two planes: an asynchronous ingestion pipeline for making videos searchable, and a synchronous serving tier for handling user queries. When a new video is uploaded, it goes through a series of steps including upload to an Amazon S3 bucket, metadata extraction, chunking of the video into segments, embedding generation, indexing, and finally serving the search results. This architecture allows for separate scaling, failure, and evolution of the ingestion and querying planes, making the system more efficient and reliable.",
  "summary": "Condé Nast's editorial teams spent an average of 250 minutes per task searching a library of more than 140,000 videos using only titles and descriptions. Working with the AWS Generative AI Innovation Center, they built a multimodal video discovery solution on Amazon Bedrock and Amazon OpenSearch Service that cut discovery time to under 2 minutes.",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}