Jukin Media Takes User-Generated Content to the Next Level with TrackIt

Jukin Media collates huge volumes of user-generated video content. It was previously relying on manual processes to find duplicate videos as well as to check for inappropriate content and licensing infringements.

The process of searching for suitable video content is a huge undertaking. “We rely on a large team trawling through videos across multiple channels, looking for compelling content that would appeal to global audiences,” explained Kris Shinn, VP of Engineering, Jukin Media. ”The challenge this creates is that often we have had the same video selected and ingested with more than one member of the team.”

With no efficient way to compare videos already in its library, Jukin Media found itself sourcing and purchasing the same video multiple times. With UGC videos used worldwide by media companies and advertisers, ensuring adherence to licensing requirements is crucial—as is assuring the appropriateness of the content being sourced. Until now, this has required a very long and laborious process.

“Previously, the team had to review each and every clip to check for the presence of celebrities or brands. They were also tasked with carefully verifying that the clips have appropriate language and content and would be suitable for all ages,” added Kris.

Having worked with TrackIt for a number of years, Jukin Media sought its advice to determine how to automate its manual, labor-intensive processes. At the same time, having an expert—but relatively small—IT team meant that any solution would have to be a managed service, reducing the amount of time required of the in-house team. Kris highlighted, “TrackIt has decades of experience designing and deploying media workflows. Its relationship with AWS meant it could advise us of a solution that was soon to be released that would perfectly address our challenges.”

As an AWS Advanced Consulting Partner, TrackIt was familiar with AWS Media2Cloud, which sets up a serverless ingest workflow to move video assets and associated metadata to the AWS Cloud. Media2Cloud lets customers set up a secure storage area for their content and use Amazon machine learning services to extract valuable media intelligence from the catalog. Media2Cloud lets users test Amazon Rekognition, Amazon Transcribe, Amazon Textract, Amazon Comprehend, and Amazon SageMaker Ground Truth against their assets, ensuring they can validate their asset management strategy before making large investments.

Jukin Media had already used Terraform as its provisioning tool and needed TrackIt’s help to implement Media2Cloud using Terraform, rather than integrating with AWS CloudFormation. The TrackIt team set to work translating the CloudFormation stack into Terraform modules. TrackIt integrated AWS Media2Cloud with the associated machine learning tools into the overall media workflow for Jukin Media.

“During the translation phase, we had to develop a number of workarounds to ensure that Terraform would work seamlessly with AWS Media2Cloud,” stated Brad Winett, President of TrackIt. “For all the resources that were previously created by the customer resources function within CloudFormation, we created equivalents within Terraform. The process was laborious and challenging but the result is a seamless workflow between the two solutions.”

With any artificial intelligence or machine learning tool, it is important to train models to ensure they are as effective as possible. Working with TrackIt, Jukin Media has worked to build a custom set of classification models that are tailored to its needs.

Media2Cloud works with Amazon Rekognition to identify objects, people, text, scenes, and activities in images and videos, as well as to detect inappropriate content. Amazon Transcribe converts speech to text to allow dialogue to be searched and analyzed. This can then be analyzed by Amazon Comprehend to identify the language of the dialogue; extract key phrases, places, people, and brands; and understand if the dialogue is positive or negative.

Jukin Media now has a much more streamlined and automated process. Thanks to AWS Media2Cloud, Jukin Media is able to avoid ingest of multiple copies of the same video, saving time and money. As it continues to work with TrackIt to train the AI solution, it will be able to automatically identify celebrities and brands.

This will be vital for avoiding licensing infringements and delivering revenue opportunities from those brands. Kris added, “TrackIt’s work with AWS to help develop our media AI/ML capabilities really accelerated our efforts to transform our media processes using cutting edge technology. Our collaboration with TrackIt has been a pleasure as TrackIt’s expertise complements our own.”