Google is bending its massive machine learning capabilities towards film making. The technology giant is currently testing a new algorithm that will allows content creators to replace their backgrounds in real time. More importantly, this technology is being tested to run on smartphones; removing the need for expensive computer setups.
Currently on trial with a small group of YouTube channels, the “Mobile Real-time Video Segmentation” uses multiple masks to separate the subject from its background. The entire system is rather technical, and is better explained on the Google Research blog.
All that really needs to be said is that the examples of the system at work look to be very good. According to the announcement, it works in videos of up to 100fps on the iPhone and 30fps on high-end Android devices.
However, it’s not entirely foolproof at the moment. There are still some rough edges that appear if a person moves to quickly, of if the lighting makes it difficult for the camera to separate objects from each other.
Still, it’s an interesting option for YouTubers who may not always want to deal with setting up a green screen.