Making 20+ Years of Content Searchable: How Prove It Media Is Using Generative AI to Work Faster and Smarter

Customer Challenge

Prove It Media is a creative, media and strategy agency with expertise in the energy sector. With nearly 30 years of experience, the agency has built an extensive library of campaigns, original video footage, photography, creative assets and research stored in Amazon S3.

That history is a tremendous resource, but finding the right content isn't always easy. As the archive grew, the team increasingly spent valuable time searching through folders, reviewing files and tracking down assets from past projects.

Prove It Media saw an opportunity to use AI to make its creative archive more accessible and useful. The goal was simple: give the team the ability to search decades of content using everyday language, find exactly what they need and spend less time looking for assets and more time putting them to work.

Implementation

TrackIt designed and implemented a contextual asset search platform that integrates directly with Prove It Media's existing AWS-based storage environment.

Frontend and Search API Solution Architecture 

The solution leverages Amazon Bedrock and TwelveLabs to analyze and understand the content of images and videos, while document assets are processed using Amazon Bedrock embedding models. Asset ingestion is automatically triggered when new content is added to Amazon S3, enabling continuous indexing and searchability across the media library.

To support semantic search, TrackIt implemented an ingestion and embedding pipeline using AWS services including Amazon SQS and AWS Lambda, with workflows orchestrated by AWS Step Functions. Generated embeddings and asset metadata are stored in an Amazon S3 Vectors bucket, creating a centralized vector index capable of supporting contextual searches across images, videos, and documents.

Ingestion and Embedding Pipeline Solution Architecture

The solution includes:

• Automated ingestion of media and document assets from Amazon S3
• Embedding generation for images, videos, and documents
• Vector-based indexing and metadata storage in Amazon S3 Vectors
• Natural language search capabilities powered by Amazon Bedrock, using TwelveLabs Marengo and Amazon Titan models
• Document summarization workflows using Amazon Bedrock Knowledge Bases
• Timestamp-based retrieval for video search results
• Secure access to source assets through generated S3 links

To provide users with an intuitive experience, TrackIt also developed a lightweight search interface where users can submit natural language queries, review search results, view asset metadata, and access matching content directly.

Outcomes

Prove It Media's team can now search for content based on meaning and context rather than relying solely on filenames, tags, or folder structures. Images, video segments, and documents can be surfaced through natural language queries, while AI-generated summaries provide additional context for document-based content. The result is less time searching for content and more time putting decades of creative work and experience to use.