Nov 16, 2026: AI_COMPLETE adds support for Twelve Labs Pegasus 1.5 (Preview)

Snowflake Cortex AI_COMPLETE now supports Twelve Labs Pegasus 1.5, a video understanding model designed to analyze and extract insights from video content. This addition enables organizations to process videos directly in Snowflake using natural language prompts, making it easier to summarize recordings, identify key events, and extract structured information from video files without building separate video processing pipelines.

Key capabilities include:

  • Advanced video understanding: Analyze video content using Pegasus 1.5 to identify scenes, actions, objects, and events across recordings.
  • Video summarization: Generate natural language summaries of video content, helping users quickly understand lengthy recordings without watching them in full.
  • Video question answering: Ask questions about video content using natural language prompts to retrieve relevant information and insights.
  • Structured information extraction: Extract information from videos for downstream analytics, reporting, and AI-powered applications.
  • Support for long-form videos: Process video files up to 2 hours in duration and 2 GB in size, enabling analysis of longer recordings.
  • Seamless integration with Snowflake: Use Pegasus 1.5 through existing AI_COMPLETE SQL workflows to analyze video files stored in Snowflake stages.

Pegasus 1.5 supports MP4, MOV, AVI, MKV, and WebM video formats. Video processing with AI_COMPLETE is currently available in public preview.

For more information, see Cortex AI Functions: Multimodal.