Topic Run shows through speech recognition?
- This topic has 2 replies, 3 voices, and was last updated 6 years, 8 months ago by
.
-
Posts
-
I’ve been playing with Google’s Cloud Speech API, and had the brilliant (silly) idea of running Loveline recordings through it. Unfortunately, the API is meant more for real-time speech processing of one speaker, but it did a surprisingly good job with the two shows I threw at it to test.
It seems to take about 30 minutes to process one show, but their quota/payment system is even more confusing than other cloud providers, so I have no idea what it might cost to pass the entire lovelinetapes archive through it.
Makes for really simple and fast metadata for a show, as well as enabling you to quickly find all instances of Drew talking about the Olsen twins.
1999-09-16: https://pastebin.com/YyZkimdT
2000-12-03: https://pastebin.com/BAKnevV0Anyone else played with this sort of thing?
This post has received 1 vote up.WOW! This is very cool. Would love to do this for all shows when the reliability is there. Not sure it’s good enough to roll this out just yet, though.
“Serendipity when I roll over my anus” (confidence=0.881339)
For a while, I was hoping users would be submitting full transcripts for shows. I have 1 or 2 of them. If we could get reliable text versions of shows, it would make the site’s search functionality a million times more awesome.
This post has received 1 vote up.
- You must be logged in to reply to this topic.