Pegasus 1.6 by TwelveLabs

What it does

Pegasus 1.6 is a video language model from TwelveLabs: you pass in a video or image and get text back, such as summaries, chapters, captions or answers to specific questions about the content. The model analyses visuals, speech, audio and on-screen text together and can split videos into individual actions or events with timestamps. According to the vendor, it handles videos up to two hours long. Access is via API and SDKs, a browser-based playground, an MCP server and AWS Bedrock.

Who it suits

Media companies, agencies, sports providers and teams with large video archives who want to describe content automatically, make it searchable or check it against guidelines. You need development skills to build it into your own workflows.

Data protection for businesses

TwelveLabs hosts its services in the US, and data may also be processed in South Korea. The vendor does not mention EU data hosting. Under the general terms of use, your data is used for training by default; you can switch this off in your account settings. A data processing agreement with EU Standard Contractual Clauses is available as an attachment to the Enterprise terms. The Enterprise plan also offers private or self-managed deployments. TwelveLabs states that it is SOC 2 Type II certified.

Limits

Billing is based on video hours and tokens, so costs are hard to estimate in advance. On the free plan, indexes expire after 90 days. The playground is mainly for trying things out; for everyday use you need your own integration. Results can contain errors and should be checked.