Twelve Labs, a multimodal video understanding AI company, said on Sept. 11 it will provide its Marengo video embedding and search model through Amazon Bedrock Managed Knowledge Base.

The managed knowledge base is a service in which AWS handles everything from collecting to indexing and searching data so AI can find and use a company’s data. Twelve Labs said Marengo is the first video model to be listed, allowing customers to search their stored videos for desired scenes using sentences without building a separate search system.

For example, entering “Show me the second-half penalty kick scene” lets users view only that segment from thousands of hours of footage.

Twelve Labs said that previously, implementing such search required extracting frames, converting speech to text and generating embeddings, then connecting them to a separate database. With the managed service replacing that process, customers can start searching by simply connecting their video repository, it said.

Danny Nicolopoulos (대니 니콜로풀로스), head of strategic partnerships at Twelve Labs, said the company eliminated the biggest obstacle that had blocked adoption of video search by working with AWS. “If your videos are on Amazon S3, you can use Marengo multimodal search right away without writing code or building vector search infrastructure. Semantic-based video search has made a major leap,” he said.