Google Announces Automatic Captions on YouTube

You are here

Google Announces Automatic Captions on YouTube

Nov 20 2008

Every minute of every day, 20 hours of videos are uploaded to YouTube. It’s hard to imagine so much uncaptioned video all in one place.

On November 19, Brenda Battat, executive director of Hearing Loss Association, and Lise Hamlin, director of public policy & state development were invited to attend the official announcement by Google and YouTube of the launching of innovative software that will make more captioning available via YouTube. Thanks to this new software, whenever a video is uploaded to YouTube, the video owner now has an option to easily add captions. The software automatically creates time-coded captions for the text of the audio quickly, easily and for free. Those captions add value to videos: videos with captions are searchable by text. That’s good not just for people with hearing loss, it’s good for anyone who wants their video to be found on Internet.

Google has also found a way for viewers to get captions on videos already uploaded to YouTube. A viewer will be able to click on a key that says “transcribe audio.” That command will add captions to videos they want to see when they want to see it. This clever technique provides one answer to the question, how anyone possibly caption every video on YouTube? It’s all done by machine. It uses voice recognition technology to automatically caption, or “auto-caption” the video.

Google and YouTube are in the “beta testing” phase with 13 educational partners. Because viewer-added captioning relies on speech recognition technology, the captions are not yet perfect. In fact, Google admitted to a 20% error rate, far below the 2% error rate we have come to expect from good caption writers on broadcast television. YouTube videos that have music or noise or environmental sounds in the background will be even more problematic for accurate automatic captions.

Still, it’s a huge leap forward for captioning on the Internet. According to Brenda Battat, executive director of Hearing Loss Association, “This is unprecedented because of the scale, it’s been done without mandates, and it’s free.”

The passion of the Google team was evident at the event in Washington DC on November 13. Ken Harrenstien, the software engineer who helped develop the automatic captioning system and who is deaf, indicated the technology has never been applied on such a large scale. “This is some thing that I have dreamt of for many years,” Mr. Harrenstien said at the event.

Someday, we do hope to see voice recognition software create more accurate captions. When the next American Idol video goes viral, we’d love to see the captioned version right from the start. In the meantime, the fact that Google and YouTube have put their weight behind captioning is a terrific development. We expect to see more and more video creators and producers understanding the value of searchable text captions that we can all enjoy.

Read more about it on Google's blog