Twelve Labs’ Marengo Added to Amazon Bedrock Knowledge Bases

Twelve Labs’ video-understanding AI model Marengo (Marengo) has been added to Amazon Bedrock Knowledge Bases. It is the first video model to be listed in Amazon Bedrock Knowledge Bases. Twelve Labs announced on the 11th that it will provide Marengo, a video embedding and search model, through Amazon Bedrock Knowledge Bases.

Amazon Bedrock Knowledge Bases is a service that helps companies use their own data with AI. The service supports the collection, indexing and retrieval process.

With Marengo added, users can now search for needed segments in natural language based on scenes, text, speech and sounds in videos.

To date, building video search has required separate processing for frame extraction, speech-to-text conversion and embedding generation. It also required connecting a separate database, and the process of integrating indexing and search results across heterogeneous information had to be designed separately.

Twelve Labs’ Marengo is configured so that when customers connect video storage such as Amazon S3 to Amazon Bedrock Knowledge Bases, it is processed within Bedrock Knowledge Bases. Once storage is connected, data collection, indexing and retrieval are linked, eliminating the need to build a separate vector database or a separate search pipeline. The search method is based on natural language.

Customers can query sports videos by entering requests such as, "Show me the penalty kick scene in the second half." Marengo searches through videos totaling several thousand hours to find the relevant segment. Searches can also be narrowed by specifying dates, categories and tags.

Marengo analyzes video scenes, on-screen text, speech and sounds. By combining visual information, dialogue and ambient sound, it prioritizes results that are highly relevant to the user's question.

The data processing method was designed with enterprise environments in mind. Video and related data are processed within the customer's AWS account, and the data does not move to an external environment. Based on this, Twelve Labs plans to expand its use cases to media, finance, healthcare and public-sector fields. These sectors have high data management requirements.

Amazon said the prerequisite for using video assets had been search infrastructure. Jose Cunaqal John, head of Amazon Agentic AI, said studios, sports leagues and broadcasters hold large amounts of valuable video, and that building search infrastructure is necessary to unlock that existing value. He added that now, users can search for desired scenes in natural language simply by specifying a video archive.

In this context, Iconik has adopted Marengo. Iconik is Backlight's cloud-based media asset management (MAM) platform. Through this integration, Twelve Labs' multimodal video search capabilities are being provided in the AWS environment.

Semantically search based on Twelve Labs is included in Iconik platform's core features. Through this partnership, Iconik has been selected as the first company in Twelve Labs' partner ecosystem to receive the top-tier "Premier Tier."

Twelve Labs plans to present the Iconik integration case at "IBC 2026," which will be held from the 11th to the 14th in Amsterdam, the Netherlands.

Dani Nikolopoulos, head of strategic partnerships at Twelve Labs, explained the conditions for using Marengo and said that if videos are stored on Amazon S3, users can use Marengo multimodal search without writing code, and can immediately use Marengo multimodal search without building vector search infrastructure. Based on this, he said semantic video search had made a major leap.

Source: TECHWORLD · Kim Seung-gi
Original: https://www.epnc.co.kr/news/articleView.html?idxno=406836

References

• https://www.epnc.co.kr/news/articleView.html?idxno=406836