×

How Machine Learning Turned Major Lazer’s ‘Make It Hot’ Into Vertical Video (EXCLUSIVE)

major lazer make it hot video
Courtesy of Major Lazer

Major Lazer & Anitta’s new track “Make It Hot” has all the makings of a summer hit — including a version that plays well on the go: Alongside the official video, artist manager services company Mtheory released a version on Wednesday that was cut for vertical viewing on Instagram and TikTok.

“With the way fans are consuming video today, we wanted to have the video live in as many places as possible, so we needed to make it in vertical format,” said Mtheory partner Zack Gershen.

But instead of manually cropping a mobile-optimized version of the clip, Mtheory used state-of-the-art machine learning to transform “Make It Hot” into a potential TikTok hit. To do this, Mtheory partnered with GoVertical, a new startup that is specializing in automating the conversion of regular landscape clips to vertical video.

GoVertical’s co-founders Irfaan Premji and John Gardiner realized the need for such a solution when they worked on their own social music app. “We wanted to incorporate music videos, but we didn’t want to force you to turn your phone,” recalled Premji.

Manually editing each and every video didn’t seem like a good option, and forcing a landscape-oriented video to show vertically didn’t work either. “You can’t just crop a video centered,” Premji said.

That’s why the duo began to develop machine learning technology that would use image recognition to identify the parts of a video worth keeping and the parts that could be safely discarded. “The system has object as well as human detection,” Gardiner explained.

Premji said that GoVertical’s system still needed a bit of hand-holding, which includes a human review to make sure the algorithm didn’t accidentally cut something essential. “Right now, it’s a fairly bespoke service,” he admitted, but added that it would improve over time.

Eventually, GoVertical also wants to develop an API as well as a self-service web interface to give customers a way to turn their own assets into vertical videos. And over time, GoVertical also wants to expand from music videos to other genres, including ads and movie trailers — assets that are often recut by hand these days, with mixed success. “A lot of these social media teams, they are not video editing experts,” said Gardiner.

Ultimately, algorithms can do a better job at converting these videos, especially at scale — which is what makes it interesting to Mtheory, a company that provides marketing and operations infrastructure for a number of artist managers. “Our friends at GoVertical did an amazing job converting it and now we’re using it on Instagram, TikTok, etc,” said Gershen. “Making videos available in vertical format is really transformational for our artist marketing.”