Amazon Bedrock has added support for TwelveLabs Pegasus 1.5, a video‑to‑text model that reads an entire video and returns structured, time‑coded metadata. Users can ask the model to locate editorial segments, speaker changes, sports plays or brand appearances and get exact timestamps. The service works through Bedrock’s US and Global cross‑Region inference profiles, with no need for a separate indexing pipeline.
Why it matters
Developers can instantly search large video libraries for specific moments, speeding up content moderation and retrieval tasks.
