# machine-learning
- a
Alphabet
03/26/2022, 3:44 AMso here's the atenborough talknet test with 10 minutes of audio and 20 minutes of audio. aten10 = 10 minutes. aten20 = 20 minutes. - a
Alphabet
03/26/2022, 3:45 AM10 minutes of audio isn't enough. There were some errors when I tried it. 20 minutes of audio seems perfect though - a
Alphabet
03/26/2022, 3:45 AMso the cutoff is somewhere inbetween 10 minutes and 20 minutes of audio - a
Alphabet
03/26/2022, 3:46 AMpretty scary that you can copy someone's voice with that little. I'm surprised fraud isn't more common with this - a
Alphabet
03/26/2022, 3:48 AMbut just 10 minutes of audio has a pretty good result too - u
(Dawn) Will Draw Fictional Women
03/26/2022, 8:47 PMno one that has to resort to fraud is smart enough to do this - h
hecko
03/26/2022, 9:14 PMconsider: there are literal kids here doing this, some with no help - h
hecko
03/26/2022, 9:14 PMall you need is intermediate computer literacy and a few hours of drudge work - u
(Dawn) Will Draw Fictional Women
03/26/2022, 9:14 PMrephrase - u
(Dawn) Will Draw Fictional Women
03/26/2022, 9:14 PMno one that has to resort to fraud is smart or willed enough to do this well - g
Gosmokeless28
03/26/2022, 9:15 PMYou said this just as I got a message from Nicholas in DMs, lol - h
hecko
03/26/2022, 9:15 PMnicholas is doing it with a lot of help - h
hecko
03/26/2022, 9:16 PMbut at least he's doing it - h
hecko
03/26/2022, 9:17 PMand it doesn't need to be done super well because phone calls have awful quality - u
(Dawn) Will Draw Fictional Women
03/26/2022, 9:17 PMuhhhh - u
(Dawn) Will Draw Fictional Women
03/26/2022, 9:17 PMlol. lmao. - s
Sweetie Belle
03/26/2022, 9:18 PM😱 - h
hecko
03/26/2022, 9:18 PMand xvatrainer is coming in a week and it'll have everything in gui - u
(Dawn) Will Draw Fictional Women
03/26/2022, 9:19 PMthe next logical step is to just make phone calls high quality - s
Sweetie Belle
03/26/2022, 9:19 PMand FASTER than real time speed - h
hecko
03/26/2022, 9:19 PMthat is a thing actually but you can just, not do it - h
hecko
03/26/2022, 9:20 PMi mean i was assuming prerecorded lines, though voice conversion would be ideal - s
Sweetie Belle
03/26/2022, 9:22 PMwow, speech to speech conversion - s
Sweetie Belle
03/26/2022, 9:42 PMmultilingual - c
clue moon
03/27/2022, 1:25 AMi think this good ...perhaps - a
Alphabet
03/27/2022, 2:22 AMoh, xvatrainer looks interesting. Still no way to copy pitch/tone like talknet though, so it's still limited in usage.