Twelve Labs

Twelve Labs builds video-native AI infrastructure that transforms raw video into searchable, AI-ready data at massive scale. The platform ingests multimodal data through a single pipeline at approximately 60x real-time speed, indexing an hour of video in a minute and supporting over 10,000 hours per day. It powers use cases including natural language video search, content compliance scanning, highlight generation, and deep video analysis through two flagship models: Marengo and Pegasus.

• Marengo (Video Foundation Model) — analyzes frames and their temporal relationships along with speech and sound, enabling state-of-the-art search and any-to-any retrieval tasks across entire video libraries.
• Pegasus (Video-Language Model) — integrates visual, audio, and speech information to generate rich text outputs including summaries, transcripts, and time-based metadata from video.
• Natural Language Search — locate specific actions, scenes, dialogue, and human emotions across hours or years of footage without manual tagging, using a single multimodal index.
• Embed API — generate embeddings from video, audio, image, and text inputs for downstream retrieval and classification tasks.
• Analyze API — extract structured insights, generate highlights, ensure compliance, and segment content from video using Pegasus.
• MCP & SDK Support — access the platform via REST API, Python/Node SDKs, and MCP integration for agent-based workflows.
• Scalable Infrastructure — supports unlimited indexing, up to 10,000-hour indexes, and 100,000 videos per index on paid plans.
• Playground — try the platform immediately without a credit card using the free tier, which includes 600 minutes of indexing.