Global Media Publisher: AI-Powered Multimodal Video Discovery on AWS
Content discovery dropped from about 250 minutes to about 2 minutes per task with Amazon Bedrock and multimodal AI.
A global premium media publisher made its library of more than 140,000 videos searchable by meaning. Creative teams describe the moment they need in plain language and get the exact clip and timestamp.
Valuable content existed. Finding the right moment was the bottleneck.
Editorial and Creative Optimization teams searched a large video archive using titles, descriptions, metadata and institutional knowledge.
- Existing search could not see inside a video. Visual, audio and transcript signals were invisible to it.
- A representative discovery task took about 250 minutes of searching and manual review.
- Assets stayed hidden whenever their metadata did not match how an editor described the content.
- The publisher needed to backfill a large historical library and keep indexing new content as it arrived.
- Enterprise deployment required secure APIs, controlled access, resilient processing, failure tracking, monitoring and repeatable operations.
Multimodal content intelligence built on AWS
The solution combines intent-based semantic search, multimodal embeddings, metadata enrichment, enterprise APIs and production-grade ingestion. Users find content by what is said, shown and heard in it.
Intent-based discovery
Users describe what they need in natural language and get contextually relevant clips with precise timestamps.
Multimodal understanding
TwelveLabs Marengo on Amazon Bedrock creates embeddings across visual, audio and transcript signals, so hard-to-find moments become searchable.
Enterprise-ready search
Amazon OpenSearch Service runs low-latency vector similarity search with metadata filtering. The surrounding AWS architecture adds resilience, security and observability.
From content discovery bottleneck to measurable operating leverage
A May 2026 benchmarking workshop measured the impact on a real Creative Optimization workflow. The results show business productivity gains on top of better search quality.
Measured task completion fell from about 250 minutes to about 2 minutes.
Teams get targeted clips with precise timestamps. Scrubbing through individual videos by hand dropped by more than 90%.
AWS reported $774,000 to $1.93 million in estimated annual operational savings, based on productivity gains from cutting repetitive manual search.
Faster discovery shortens response time to advertiser, sales and brand-partnership requests and increases reuse of existing content.
The benchmark used a representative workflow from the customer's 13-member Creative Optimization team. The results set a baseline for similar discovery workflows across other editorial and commercial teams. Broader adoption is expected to add productivity gains and help teams find content for advertiser and brand-partnership opportunities faster.
Built to operate across a growing enterprise content library
The platform separates compute-heavy ingestion from low-latency search, so each plane can scale, fail and evolve independently.
Technical proof points
- About 150,000 historical records processed in 4.5 to 5 days using eight concurrent requests.
- Historical backfill and incremental ingestion run in the same operating model.
- Retries, reprocessing, failure categorization and traceability are built into ingestion operations.
- Secure versioned APIs with authentication, rate limiting, caching, error handling and monitoring.
- Infrastructure-as-code documentation supports repeatable deployment and ongoing platform management.
AWS published this solution and its measured results
The AWS Machine Learning Blog covered the implementation on 29 September 2026 in "How Condé Nast built multimodal video discovery with Amazon Bedrock." The post documents the architecture, the multimodal search capabilities and the results, including the 99.2% cut in discovery time and $774,000 to $1.93 million in estimated annual operational savings.
Modernize content discovery with multimodal AI on AWS
Build a secure, scalable content intelligence foundation so teams find and reuse high-value assets faster.
Book a Demo