Speech Recognition
Related
Machine Learning
AKA
<<<ASR>>>
Automatic Speech Recognition
<<<STT>>>
Speech To Text
KaldiASR
ossSupports offline
Models for english
GitHub
Mozilla's DeepSpeech
ossSupports offline
Models for english
GitHub
Examples
Results for 1st minute [ from https://www.youtube.com/watch?v=-I_jE0l7sYQ ]
it tatatatatat young i started in two thousand iterating to mash it in
london i say i think to moses fantasca change is the names of the complices
every year men and so when when you're referring to the previous
confinement quite remember what i would call and if he read my glove you
should really now you consented essay the one check some of the title of
her and then of the mutable and underways givin it in the future then way
this is as comes he nerto thousand thirteen ten and francesca begod cook
actually figured out now that francesca leap to the mean cook that was
because he was a
Computation time
~1m
Seems to scale linearly with input. ~30s of input takes ~30s of processing
OpenAI's Whisper
ossOn GH
CPU-powered fork
So far is a great, open-source (MIT licensed), Speech Recognition engine which is easy to install and use, and yields great results in reasonable times.
Usage
Results for (49min) of https://www.youtube.com/watch?v=-I_jE0l7sYQ , tiny model
whisper test.m4a --language English --model tiny
Times
real 34m57,850s
user 62m10,509s
sys 0m57,646s
[00:00.000 --> 00:26.000] The started in 2013. I think code mesh in London. I say I think code mesh because Francesco
[00:26.000 --> 00:33.520] changes the names of the conferences every year. When you're referring to a previous
[00:33.520 --> 00:39.280] conference, you can never remember what it was called. If you read my blog, you should
[00:39.280 --> 00:44.360] really name the conferences by the SHH1 check some of the title. Then it would be a
[00:44.360 --> 00:49.520] mutable and you'd always be able to find it in the future. Anyway, this was code mesh
[00:49.520 --> 00:54.120] in 2013. Francesco is a very good cook.
[00:54.120 --> 01:00.440] Actually, the thing you don't know about Francesco is he's a mean cook because he was a short
[01:00.440 --> 01:07.400] order cook in the university. I mean, he was certainly a bit of extra money in his spare time.