After bringing automated captions to YouTube back in 2009, the service has been working on improving the accuracy of the feature. And now according to YouTube’s product manager Liat Kaver, there are now over a billion videos that have auto captions on them, thus giving access to those with hearing impairments as well as those who don’t speak the language spoken on the videos. Since the launch, Kaver says they have achieved a 50 percent leap in accuracy when it comes to the English captions thanks to the improvements of the service’s machine learning algorithms and expanding its training data. You can see the difference in the photo above between the old service (left) and the current model (right).
Going forward, Kaver, who writes about her own hearing disability in this announcement, says they plan to improve accuracy for 10 supported languages. They’re calling on creators to help and review the accuracy of these machine-generated captions to help improve the technology quicker.