What’s next in video tech: AI, avatars & more

Welcome to Lowpass! This week: How AI, avatars and other cutting-edge tech can transform video, and Humane’s maybe-pivot.

Last week, hundreds of video engineers flocked to San Francisco to discuss and learn about the latest in video technology at the industry’s annual Demuxed conference. Speakers at the two-day event talked about film grain synthesis, CNDs, content encryption and video processing on quantum computers – geeky stuff, presented by insiders to insiders, without the heavy-handed sales pitches that dominate so many other industry events.

I had the chance to attend one day of Demuxed in person, and tuned into some panels remotely the other day. And while I have to admit that some of the content went way over my head – I’m not a video engineer, after all – I also truly enjoyed the collaborative atmosphere, the Big Buck Bunny jokes, and the behind-the-scenes peeks at how some of the biggest companies in media are using cutting edge video technology to improve everyone’s streaming experience.

A few talks that stood out to me in particular for outlining novel solutions to widespread industry challenges:

In live video, you gotta fake it till you make it. Netflix is known for both embracing cutting-edge tech, and obsessive testing to make sure that technology delivers on what it promises. One example for this is the company’s video encoding work, which includes optimizing the encoding settings for each and every shot of a movie or TV show episode. I wrote a deep dive about this for The Verge earlier this year, which also noted that the company has been taking a fairly conservative approach when it comes to optimizing live video streams.

“We’re quite early into live streaming,” Netflix’s senior encoding technology director Anne Aaron told me at the time, acknowledging that the primary goal was to build a scalable and resilient live streaming infrastructure. “We’re starting with a bit more industry-standard ways to do it,” she told me for my article. “And then from there, we’ll optimize.”

Last week at Demuxed, members of Aaron’s team explained one key challenge they’re facing with Netflix’s still-nascent live programming. “We have many more ideas than live events,” said Netflix staff software engineer Wei Wei. A/B-testing each of those ideas with a live event would simply take too long, and running experiments that could go wrong during something as high-profile as a NFL live Christmas game isn’t the best way to build out a live business either.

Subscribe to Premium to read the rest.