Introduction

Kubrick is an open-source end-to-end ingestion pipeline and API for video semantic search. It allows users to search video using natural language, images, audio, or other video clips, returning the exact moments that are most relevant.

As video libraries grow across domains like sports, education, security, and marketing, finding a specific moment becomes tedious. Scrubbing through timelines, relying on inconsistent tags, or remembering file names is slow and often frustrating.

Building a system that can understand video content is complex. It requires processing large files, extracting multimodal features, storing and indexing them efficiently, and serving fast search results, all while integrating into existing infrastructure.

Kubrick’s pipeline automatically generates vector embeddings from ingested video and stores them in a vector database. The API allows users to search the database and interact with the pipeline. Its modular, serverless architecture can be deployed directly into an AWS environment, giving teams full ownership of their data while remaining flexible enough to integrate new embedding models in the future.