How Social Department built culture-aware video discovery
Social Department, an AI operating system for entertainment marketing, worked with TrackIt and AWS to build QuickSearch: single-action queries that surface shareable scenes, quotable lines, memeable frames, and trend-aligned moments across large video libraries, using TwelveLabs models on Amazon Bedrock and Amazon OpenSearch Serverless.
This post, co-written by Mariah Miller and Sirajus Salekin from AWS and Nicholas Chase from Social Department, discusses the creation of QuickSearch, an AI-operated system designed to simplify video discovery for entertainment marketing teams. Entertainment teams manage vast amounts of video assets, but traditional search methods often fall short in finding the most relevant clips for campaigns.
QuickSearch addresses this issue by encoding editorial questions as executable retrieval pipelines. The system offers four actions: Shareable Scenes, Quotable Quotes, Memeable Moments, and Trend Align. Each action employs its own retrieval strategy, editorial criteria, and ranking model, allowing marketers to find the right video moments without needing to describe them explicitly.
Shareable Scenes identifies substantial, self-contained scenes like extended confrontations, climaxes, and set pieces, while Quotable Quotes focuses on punchy, quotable dialogue, even if those clips are short. Memeable Moments detects individual frames suitable for memes and generates captions based on the content of those frames. Trend Align takes into account trends happening outside the video library, ensuring the returned clips are relevant to current audience discussions.
To ensure responsible AI usage, generated outputs are checked against the source material, and every post is approved by a member of the customer's team before publishing. The system leverages Amazon Bedrock for model access, and TrackIt, an AWS Premier Tier Services Partner, implemented and delivered the system design. The solution uses Twelve Labs' Marengo 3.0 and Pegasus 1.2 models for scene-level video reasoning and Amazon OpenSearch Service with Amazon OpenSearch Serverless for retrieval.
Written by urgent.news from AWS Machine Learning's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.