Recently I noticed another feature added in Youtube, largest online video server to transcribe the audios of the videos. This is based on Google's voice recognition.
You can notice a button marked CC at the bottom of the player as shown in the picture. Clock on this and select transcribe audio beta, you will be able to see the captions at the bottom as you see in movies. There are quite a many language options are also there under translate captions I have not tested. The above picture is taken from the movie Before Sunrise. I have observed a lot many mistakes in the captions. Youtube tells this feature is under test.
But this transcribe feature is far from perfect yet. More than 75% of the texts are wrong and funny. It gives some unrelated word as captions and makes it funny. Many a times it just gives some random words which are really funny. You can also experience it once. Click on the picture above to see it enlarged and you will get to know how to get the captions.