53 ms·
Voice recognition is astoundingly fantastic even on poor hardware. If you want to see the state of the art try dictating a few lines here - https://speechnotes
by logicallee 10y ago
Voice recognition is astoundingly fantastic even on poor hardware. If you want to see the state of the art try dictating a few lines here - https://speechnotes.co/ https://speechnotes.co/ no registration required, no training required, use any microphone you want.
I was absolutely astounded when I found out the state of the art today. I was used to, "Computer, how do you wreck a nice beach" from a couple of decades ago.
Why can't a place like Stanford auto-caption things? Especially things like lectures which are super easy - a single speaker droning on in a silent room using high-quality speakers. without any editing the output should be absolutely useable.
The only time it really wouldn't work at all is in a very noisy environment where you can hardly understand it anyway.
I realize some of the captions might end up garbage, but I don't think enough would to make it into a complaint.
- detaro 10y agoYour link uses Google APIs, nothing local, so it would be weird if Youtube (which apparently isn't good enough) used a worse version of those. But who knows...
- logicallee 10y agoHow did you find my link? (I mean, what did you think of it, assuming you tried it). By the way, I find Google's autogenerated CC is useless on any video no matter the type or audio quality. (For example a lecture.) Another reason to think the link uses something else is that it takes several seconds (during pauses) to come up with the result - it is slower than real-time. I don't know if Google takes the time to slowly process every video in case someone clicks the autogenerated CC... The quality definitely strikes me as different. (And one of the reasons I was so surprised that the above link is actually useable.)
- Nimitz14 10y agospeechmatics.com