AI model for transforming video into Time-Based Metadata
AI that truly understands video. Uses multimodal models (Marengo/Pegasus) to search, analyze & generate text from video content at scale.
Pegasus 1.5 understands narrative structure and pacing to automatically segment content, identify natural breaks, and assemble rough cuts from hundreds of hours of footage. Users describe desired output and the system exports directly into editing workflows.
Pegasus 1.5 ingests video, audio, and image data at ~60x real-time speed, indexing one hour of video in one minute. The unified index supports search, analysis, and embedding tasks across all modalities without separate processing pipelines.
Pegasus 1.5 reasons continuously over full video duration to identify patterns and signals, enabling teams to understand what content performs well and make data-driven creative decisions at scale.
Pegasus 1.5 enables semantic search across entire video libraries by understanding visual content, dialogue, emotions, and actions without manual tagging. Users describe what they need (e.g., 'all scored goals this season') and the system returns timestamped clips, reducing research time from days to seconds.
Pegasus 1.5 analyzes video content to detect policy violations and brand safety issues with explainable AI, allowing compliance teams to review findings faster and with confidence in automated decisions. This accelerates content moderation workflows by 10x.