Marengo
#12 in KI-Video-EditingTwelvelabs · 2× · tolest 15. Sept. 2026
43
Momentum
Marengo is a proprietary multimodal video-embedding model by TwelveLabs that analyzes visual, acoustic, linguistic, and motion information in videos to enable "Any-to-Any" search (text, image, audio, video). The current version Marengo 3.0 was launched on December 1, 2025 at AWS re:Invent as a General-Availability release and generates compact 512-dimensional vector embeddings for videos up to four hours long in 36 languages. The model is available via the native TwelveLabs API/platform as well as as a managed model through Amazon Bedrock (including Bedrock Managed Knowledge Base) and is designed for enterprise applications such as semantic video search, anomaly detection, and RAG systems.
Momentum-Verloop
18.06.16.09.
Features
| Output Formats | 512-dimensional vector embeddings (JSON), incl. segment start/end times for video moments |
| Base Model | Standalone, video-native multimodal foundation model (not adapted from image models), current version Marengo 3.0 |
| Integrations | Amazon Bedrock (incl. Managed Knowledge Base, S3 Vectors), OpenSearch, Elasticsearch; API-first design with Embed API and Search API |
| Collaboration | No specific collaboration features documented; API-based integration into existing team/enterprise workflows (Playground for individual users) |
| License | Proprietary, commercial model; non-exclusive, non-transferable, revocable usage license per TwelveLabs Terms of Use |
| Platform | TwelveLabs API/Playground as well as Amazon Bedrock (incl. Bedrock Managed Knowledge Base) |
| Price | Video indexing from $0.042/min (Visual+Audio); search usage $4/1,000 queries; free plan up to 10 hours indexing; Enterprise custom pricing |
| Release Date | December 1, 2025 (Marengo 3.0, General Availability, announced at AWS re:Invent) |