TwelveLabs Pegasus 1.5 Arrives on Amazon Bedrock for Video Understanding
Amazon Bedrock now supports TwelveLabs Pegasus 1.5, which converts video into structured, time-coded descriptions without a separate indexing pipeline.
AWS made TwelveLabs Pegasus 1.5 available on Amazon Bedrock on October 9, 2026. The video-to-text model generates structured metadata with timestamps.
Rather than only producing a general summary, Pegasus 1.5 is designed to identify when specified events occur in footage.
The model analyzes visual content, audio and on-screen text in a single pass. Example targets include editorial segments, speaker changes, sports plays and brand appearances.
AWS emphasizes that a separate pre-indexing or ingestion pipeline is not required before analysis. Large-scale production systems may still need their own storage and retrieval architecture.
Media teams could use time-coded metadata to find relevant footage more efficiently. Human editors still need to verify candidate clips.
Sports and advertising analytics are additional potential uses, but identity, copyright and usage rights are separate issues that a model output cannot resolve.
AWS lists US and Global cross-region inference profiles for access. Data residency and contractual restrictions should be reviewed before sending sensitive footage.
Organizations should benchmark detection accuracy, timestamp precision and cost using representative video rather than relying on a single demonstration.