Twelve Labs

Twelve Labs builds multimodal AI models that let developers and enterprises search, analyze, and extract insights from video at scale, combining visual, audio, and language understanding rather than relying on transcripts alone. The company's two core models are Marengo, a multimodal embedding model supporting 47 languages, and Pegasus, a video-language model built to reason over full video sequences up to two hours long rather than just reading captions. With these models, users can run natural-language searches across large video libraries to find specific actions, scenes, dialogue, or emotional moments, automatically segment and organize footage, generate highlight reels, and flag anomalies for compliance and content-moderation use cases. The company states it can index roughly one hour of video per minute and processes more than 10,000 hours of video daily across its customers. Access is available through a web-based playground for testing, a full API and SDKs for production integration, and Model Context Protocol support for agent-based workflows. The platform is SOC 2 Type II certified and offers on-premises deployment for security-sensitive customers. Twelve Labs was founded in 2021 and serves media, sports, advertising, government, and security organizations, including partnerships with NFL Media and MLSE. Pricing is not published; prospective customers must contact sales for quotes.