machine learning

Automated Content Moderation with AWS Machine Learning

The AWS Rekognition POC enables companies the ability to moderate user-generated videos using machine learning. Designed by Metal Toad and Mux. Learn more here.

Metal Toad was recently announced as a solutions partner in Mux’s newest partner program initiative. Mux is a video platform that enables developers to quickly and easily build out amazing video experiences. 

We are very excited to be partnering with Mux to bring innovative Amazon Web Services (AWS) cloud solutions to their platform and users. Our first opportunity to work with Mux was to build an AWS Rekognition proof-of-concept (POC) that could be integrated into a video workflow using Mux.

In this post, we’ll cover:

  • Content Moderation Opportunity 
  • Developing a Solution Using AWS
  • How the AWS Rekognition POC Works

Content Moderation Opportunity

Content moderation is an important topic to any publisher of online media. Mux and Metal Toad worked together to identify an option for Mux’s customers to incorporate content moderation on top of AWS. This allows video publishers to quickly identify potential obscenities in the videos, including violence, drugs, and alcohol. Alternately, if users took it upon themselves to manually moderate their content, this process could take hours, days, or even weeks, depending on the number of videos. This cumbersome process could result in videos that may not be appropriate to be posted on users’ websites, applications, or other platforms.

Developing a Solution Using AWS

Mux is an AWS technology partner, and many of Mux's customers build on AWS. We are also an Advanced Consulting Partner with AWS, which means we have a deep understanding of how to use and deploy AWS products and services. Leveraging our existing partnership with AWS, we decided to build an integration that would upload videos to Mux’s platform using AWS products.

Within AWS, there are multiple machine learning cloud-based software products available. These products include Amazon Sagemaker, Amazon Comprehend, and Amazon Rekognition, to name a few. In this proof of concept, we decided to go with Amazon Rekognition. Rekognition was chosen because it already has a model trained for content moderation. This means that we could integrate and deploy Rekognition faster than other machine learning software provided by AWS. With the plan in place, we worked internally and with the team at Mux to bring this proof of concept to life.

How the AWS Rekognition POC Works

The workflow is pretty simple. Let’s follow a user as we decide to upload a new video:

  1. The user uses a React site to upload a new video. The user requests a signed URL from an API so that the video can be securely uploaded to Mux.
  2. The React site then uploads the video to Mux for processing.
  3. The file upload to S3 triggers a Lambda to process the video.
  4. The Lambda orchestrates SNS, SQS, and AWS Rekognition pipeline.
  5. Once the Rekognition finishes, the data gets inserted into DynamoDB, which is viewed from the React Site.

Here’s a visual representation of the workflow outlined above.

An architectural diagram that shows how the AWS Rekognition PoC works.

We're excited to be partnering with Mux and to share this AWS Rekognition POC. If you think this is something that could work for your business or if you have any questions, please reach out to us at

Similar posts

Get notified on new marketing insights

Be the first to know about new B2B SaaS Marketing insights to build or refine your marketing function with the tools and knowledge of today’s industry.