Pegasus 1.5 video model on Amazon Bedrock

Amazon Bedrock adds TwelveLabs video model Pegasus 1.5

Short answer

TwelveLabs' Pegasus 1.5 video language model is now available on Amazon Bedrock. You define the moments you care about, and it returns every matching segment as timestamped JSON. It accepts videos up to two hours long, needs no pre-indexing, and is available in the US West (Oregon) Region.

What this means for operators

For a 10-200 person B2B company, this matters only if you hold long video: recorded demos, support sessions, training material or webinars. The source's own examples come from media and sports, so any fit with sales or support recordings is our reading, not the vendor's claim. If you do have such a library, you can define a segment once, such as a pricing discussion or a product shown on screen, and get consistent JSON for every file. That feeds a CRM, a ticket queue or a search index without anyone scrubbing through footage. If you have no long video, there is nothing here for you to act on. Check the Region and the separate AWS Marketplace billing before you plan around it.

AWS announced on Oct 9, 2026 that TwelveLabs' Pegasus 1.5 is available on Amazon Bedrock, according to the AWS Machine Learning Blog. It is a video language model that converts long-form video into structured, timestamped metadata.

What changed

Earlier Pegasus versions answered a question about a clip. Pegasus 1.5 goes through the whole video, finds every segment that fits criteria you define, and returns them as JSON with start and end times and the fields you asked for. AWS lists four changes from previous versions:

Time-based metadata. You define segments, events or cues, and the model returns each match.
Multimodal prompting. A segment definition can pair a text description with a reference image, for example to locate a specific product or logo. One request can carry up to 30 reference images.
On-the-fly processing. Videos are sent directly from Amazon S3 or as a base64 string, with no ingestion pipeline or pre-indexing. AWS says the data stays within the AWS data boundary.
Longer input. Videos up to two hours in a single request, up from one hour in Pegasus 1.2.

Limits and setup

Files can be up to 2 GB from S3, or up to 25 MB as a base64 string. Resolution runs from 360×360 to 5184×2160, and aspect ratios from 1:1 to 2.4:1. English is fully supported. Several other languages, including Russian, Spanish and German, are supported only partially.

You call the model through the InvokeModel API, or InvokeModelWithResponseStream for streamed output. The model ID is twelvelabs.pegasus-1-5-v1:0. Access has to be enabled in the Bedrock console, and your IAM policy needs the invoke permissions. Cost can be tracked per application with application inference profiles, and usage metrics appear in CloudWatch.

Two modes exist. General mode takes a text prompt and returns text, or JSON if you supply a schema. Time-based metadata mode takes segment definitions instead of a prompt and returns a list of matching segments. Several definitions can run in one request, and each returned segment is tagged with the definition it matched.

Availability and billing

The model is available in US West (Oregon). AWS points readers to its Region list for future additions. Billing runs through AWS Marketplace, so charges show up on the AWS bill and in Cost Explorer under TwelveLabs, not under Amazon Bedrock. Pricing is on the Bedrock pricing page; the announcement quotes no figures.

Intended use

AWS positions the model for organisations with large volumes of long-form video: news archives, sports broadcasters and streaming services. It says people stay in the loop, shifting from tagging footage by hand to reviewing the model's metadata.