This ebook introduces Pegasus 1.5, a video reasoning model built for time-based metadata extraction, and shows how it turns an unsearchable video archive into a structured, queryable data asset. It covers how to define custom segment schemas in plain language, process up to two hours of video in a single API call with no ingestion pipeline, and use multimodal prompting to track specific visual events like logos or players. Includes benchmark results (30%+ more accurate segmentation than Gemini 2.0 Pro) and use cases across sports analytics, media, content ops, and brand monitoring.