Pegasus 1.5 from TwelveLabs is now available in Amazon Bedrock

Summary by forck

Amazon made TwelveLabs' Pegasus 1.5 video language model available in Amazon Bedrock, letting customers turn long-form video into structured, timestamped metadata without building an ingestion pipeline.

• The model analyzes an entire video, finds every segment matching user-defined criteria, and returns each match as JSON with start and end times and the requested fields.
• It adds video segmentation, multimodal prompting with reference images, and on-the-fly processing, and it handles videos up to two hours long in a single request, up from one hour in Pegasus 1.2.
• Videos can be sent from an Amazon S3 URI or as a base64 string of up to 25 MB, with no pre-indexing required, and Amazon says the data remains within the AWS data boundary.
• Customers call it on the bedrock-runtime endpoint with the InvokeModel API or InvokeModelWithResponseStream, and can tag an application inference profile to track cost by application.
• Pegasus 1.5 keeps the earlier model's video-to-text generation, temporal grounding and multimodal understanding, with full English support and partial support for twelve other languages.