Easy Live Video using AI/ML on Top of AWS Elemental MediaLive

Easy Live Video using AI/ML
July 2, 2021

With the increased usage of video streaming for work purposes, there is a strong need to monitor live and recorded video broadcasts. The quality checks could be as simple as signal errors, issues with subtitles, and audio language, which human operators typically monitor. It becomes very difficult sometimes for live broadcasts.

The simple usage of artificial intelligence (AI) can automate many of the monitoring tasks done by human resources. The AI-based detections can help you to analyze the content of an HTTP Live Streaming (HLS) video stream. AWS reckognition performs an example set of monitoring checks in near real-time (<15 seconds).

Below is an attempt to familiarize you with the technologies and standards used in this solution –

  1. HLS is an HTTP adaptive bitrate streaming communications protocol.
  2. AWS Elemental MediaLive is a real-time video service that lets you create live outputs for broadcast and streaming.
  3. Amazon Rekognition Custom Labels allows you to build models to identify the objects and scenes specific to your business needs.

A robust broadcast quality control solution should monitor various aspects of the live streams:

A robust broadcast quality control solution should definitely monitor various aspects of the livestreams:

Traditional image and audio analysis algorithms can be used for some situations, and many are better suited for detection using Machine Learning (ML) models.

Traditional image and audio analysis algorithms
Traditional image and audio analysis algorithms


  1. The video ingestion pipeline produces HLS streams using AWS Elemental MediaLive and is stored in Amazon Simple Storage Service (Amazon S3)
  2. A video processing pipeline orchestrated by AWS Step Functions based automation which performs monitoring checks on extracted frames and audio from every video segment

Automated Verification Outcomes:

  1. Audio silence detection – based on a predefined volume threshold.
  2. Logo verification – Known logos from images are well suited for Convolutional Neural Networks (CNN) based ML models. Object detection models have been created using Amazon Rekognition Custom Labels.
  3. Program type verification: Whether the video looks like the type of program it should be. To ensure this, a custom image classification model was created using Amazon Rekognition Custom Labels.
  4. Character/person identification: Whether this video shows the correct person or actor. For verifying this, the face image extraction feature of Amazon Rekognition to look for persons/ actors on screen can be leveraged along with Rekognition Custom Labels to train a model to recognize a specific show/ program.

In combination with Media Live and Amazon AI/ML services, you can easily broadcast the content of your dream. The near real-time AI/ML intelligence would bring more efficiency to expect accuracy up to your need. Stay tuned for more media solutions using new-age technologies.

Written By,

Abhinav Abhishek

Solutions Director, Rapyder Cloud Solutions


Want to know? Contact us now for a free consultation


Cloud computing Insights and Resources

Cloud Computing Insights and Resources

gen ai evolution of cloud computing

Transforming Web Development: The Modern Era Advancement of Microservices Architecture

The evolution of traditional processes has been propelled by advancing technology and the integration of recent discoveries. The web development […]

gen ai evolution of cloud computing

Future-Proofing Infrastructure: How Generative AI Shapes the Evolution of Cloud Computing 

The reality of cloud computing stands at the edge of a vast transformation, all thanks to the emergence of Generative […]

cloud computing security

What are the security issues in cloud? 

What is cloud?  Cloud computing refers to the delivery of computing services, including servers, storage, databases, networking, software, analytics, and […]