Google has been experimenting with the automated captions for several months with a handful of high profile partners like the University of California, Berkeley, Yale University and National Geographic. All other captions on YouTube videos were provided by the videos' producers, dpa reported.
Google has been working on speech recognition technology for some five years, and uses the technology to transcribe audio voice mails through its Google Voice service, and to provide spoken Web searches from smart phones.
However, engineers warned that the technology is far from perfect and that the machine translations are sure to contain mistakes.
"We know it's not perfect, and sometimes it will be funny," said Google engineer Ken Harrenstien, who is deaf. "But it's better than nothing."