AWS Developed an AI Solution for Natural Language Video Querying

Amazon Web Services announced a new artificial intelligence solution that enables video files to be queried using natural language, utilizing the Strands Agents SDK and various cloud services.

New Era in Video Analysis

Accessing large amounts of accumulated video data in sectors such as media, security, and insurance is quite cumbersome with traditional methods. The new system allows users to ask questions about uploaded videos in natural language and receive answers within seconds.

Artificial Intelligence Architecture

The developed solution uses a dynamic structure that decides at runtime which AWS services to invoke. The system coordinates Amazon Bedrock, Amazon Rekognition, and Amazon Transcribe services through the Strands Agents SDK.

Execution Times and Performance

While responses for pre-analyzed content return in less than a second, the initial analysis of new videos takes 5 to 10 minutes. The AI agent resolves user intent, decides on visual or audio analysis, and provides timestamped responses.

Real-World Usage

A leading media and entertainment company has adopted this approach as part of the AWS Professional Services process. Manual review time in the accumulated archive consisting of multi-hour recordings has been reduced by approximately 80 percent.