Pegasus 1.5 by TwelveLabs

Pegasus 1.5 is an AI model that converts raw video streams into structured, time‑based metadata. It extracts information from visual, audio, and language cues within the footage, producing a searchable representation of events, actions, dialogue, and emotions aligned to timestamps.

The model is intended for developers and organizations that process large volumes of video and need to turn passive recordings into data that can be queried programmatically. It is accessed through TwelveLabs’ API and SDK, fitting into pipelines that ingest multimodal content at high speed.

What distinguishes Pegasus 1.5 is its ability to index an hour of video in roughly a minute, achieving about sixty times real‑time processing and handling tens of thousands of hours per day. The resulting metadata enables natural‑language search, automatic scene segmentation, compliance checks, and highlight generation without requiring pre‑existing tags.