A note on your comment "...Emily's, on the other hand, was dropped a couple of times."
I used to work for a speech technology company (developer), and our speech recognition software always struggled with my soft voice (I'm a guy!) Our brains are still a lot better at voice recognition than computers.
The timing of the captions is off in in several places the second half and some captions are bunched together, while some captions don't appear at all. Otherwise, yes ... great humor!
Yes. This is something we wanted to demonstrate: Though while YouTube Automatic Timing is a great service that will do much for getting YouTube video captioned, it is not perfect.
For some reason, my voice was aligned perfectly (or almost perfectly) throughout the short video. Emily's, on the other hand, was dropped a couple of times.
Not perfect on YouTube is great ... not accurate in an academic environment gets dicey, but I have to say that I am impressed with single speakers who speak well. The ability to get the timecoded script back and tinker with it may represent the best of both worlds. I can imagine that this is the solution for single speaker prof podcasts.
Hi Ken,
This is a very useful video - thanks!
A note on your comment "...Emily's, on the other hand, was dropped a couple of times."
I used to work for a speech technology company (developer), and our speech recognition software always struggled with my soft voice (I'm a guy!) Our brains are still a lot better at voice recognition than computers.
Cheers.
NickFreear 1 year ago
The timing of the captions is off in in several places the second half and some captions are bunched together, while some captions don't appear at all. Otherwise, yes ... great humor!
hermanchas 1 year ago
Hi hermanchas,
Yes. This is something we wanted to demonstrate: Though while YouTube Automatic Timing is a great service that will do much for getting YouTube video captioned, it is not perfect.
For some reason, my voice was aligned perfectly (or almost perfectly) throughout the short video. Emily's, on the other hand, was dropped a couple of times.
Thanks for the comment!
-ken
kennpetri 1 year ago
Not perfect on YouTube is great ... not accurate in an academic environment gets dicey, but I have to say that I am impressed with single speakers who speak well. The ability to get the timecoded script back and tinker with it may represent the best of both worlds. I can imagine that this is the solution for single speaker prof podcasts.
hermanchas 1 year ago
I found the video helpful. I liked the way they tried to add typical interference to test the accuracy. It gave me more confidence in the software.
suetunn 1 year ago
It works surprisingly well.
This was a useful test, well thought-out. You've saved me the hassle of doing a similar test myself.
I think you've just convinced me to use the automated captioning system on my latest video, for the lyrics to a Christmas carol parody song.
Good job. I like the understated humour in this video.
ROBwithaB 2 years ago