It’s about using our technology and research to aid enrich people’s lives. Much like YouTube — and its mission to give everyone a voice and show them the world.
Our collaboration with YouTube’s product and engineering teams has helped streamline decision-making, raise security and engagement, and improve experiences for all types of users.
Making it easier to find Tiny videos
YouTube brief videos less than a minute long are viewed over 50 billion times a day.
Covering everything from rising K-pop stars to local food guides, they’re quick to watch, quick to make — and increasingly popular. But because Shorts are created in minutes, they often lack the descriptions and titles that make them basic to find. That’s why we introduced Flamingo, our visual language model that helps generate descriptions.
Flamingo analyzes the opening frames of a video and explains what’s on screen (e.g., “dog balancing a stack of crackers on its head”). It saves that text as metadata on YouTube, creating clearer categories for content and matching user searches to better results.
YouTube is rolling out this technology to Shorts, with automatically generated video descriptions for all novel uploads. Now viewers can find and watch more relevant videos from a more diverse group of global creators.
Video compression optimization
Video technology has evolved rapidly in recent years, and given that internet traffic is only expected to raise in the future, video compression is becoming an increasingly pressing issue.
We worked with YouTube to test the potential of our AI model, MuZero, to improve the VP9 codec, an encoding format that helps compress and transmit video over the internet. We then applied MuZero to some of YouTube’s live traffic.
At launch, we saw an average 4% reduction in bitrate across the video sets. Bitrate helps determine the processing power and bandwidth needed to play and store videos—affecting everything from load times to resolution, buffering, and data usage.
By improving the VP9 codec on YouTube, we’ve helped reduce internet traffic, data usage, and the time it takes to load videos. And by optimizing video compression, millions of people around the world can watch more videos while using less data.
Brand Security Protection
Since 2018, our YouTube partnership has helped creators understand what types of videos can generate ad revenue and make sure the right ads are displayed in the right place.
We developed the Label Quality Model (LQM) with the YouTube team to more accurately label videos and in line with YouTube’s advertiser-friendly guidelines. In addition to improving the accuracy of ads served in videos, it helps ensure that ads are served alongside content that adheres to YouTube’s guidelines.
By improving the way we identify and classify videos, we’ve increased trust in the platform among viewers, creators, and advertisers.
Improving AutoChapters
As the way we create and watch video evolves, creators have begun adding chapters to their videos. This makes it easier for their audience to find the content they need — but it can be a ponderous process.
We worked with the YouTube Search team to develop an AI system that suggests chapter segments and titles for YouTube creators by automatically processing video transcripts, audio, and visual features. With AutoChapters, viewers spend less time searching for content and creators save time creating chapters for their videos.
Since this feature was introduced in Google I/O in 2022Auto-generated chapters have been applied to tens of millions of videos (and counting) on YouTube.
Developing technologies and products
We’re constantly looking for ways to improve Alphabet products with our AI research.
Our partnership with YouTube has already had a huge impact on people’s lives. We are working on more projects to continually improve the user experience.