# machine-learning
- u
(Dawn) Will Draw Fictional Women
02/07/2022, 9:17 PMi am also gtx 1650 haver 😎 - u
(Dawn) Will Draw Fictional Women
02/07/2022, 9:17 PM< (look at this loser does not train locally) - u
user
02/07/2022, 9:20 PMgtx 1060 to 1650 isn't enough of an upgrade for me to justify spending $300 - e
Emoto
02/07/2022, 10:04 PMhttps://librivox.org/ dropping this in for anyone who's interested in using these wavs for tacotron tts models. most if not all recordings they do are in the free domain and technically being able to be used in tts 🤔 - g
Gosmokeless28
02/07/2022, 10:14 PMThere is also https://research.google/tools/datasets/libri-tts - e
Emoto
02/07/2022, 10:16 PMspicy - e
Emoto
02/07/2022, 10:16 PM585 hours of speech would make one hell of a vocoder model - e
Emoto
02/07/2022, 10:17 PMnot sure if its good for tacotron though, i never seen a decent mixed speaker tacotron2 model - m
mega b
02/07/2022, 10:33 PM - g
Gosmokeless28
02/07/2022, 10:34 PMWhat does this do? - m
mega b
02/07/2022, 10:35 PMIt's a dataset - e
Emoto
02/07/2022, 10:36 PMoh interesting - e
Emoto
02/07/2022, 10:37 PM33k hours of semi supervised data and 10k hours of supervised data - e
Emoto
02/07/2022, 10:37 PMyou could make massive vocoder models or decent enoguh male/female models for uberduck with that - u
(Dawn) Will Draw Fictional Women
02/07/2022, 10:38 PMthat is such a fucking awesome name - c
Couch
02/07/2022, 10:38 PMGigach - e
Emoto
02/07/2022, 10:52 PMI kinda hope that uberduck will be able to use these to create more prebuild models to use except the LJSpeech one - c
Couch
02/07/2022, 10:53 PMit dont want to learn attention - e
Emoto
02/07/2022, 11:06 PMWeird - a
Awesome Face (GTX 1660 Ti)
02/07/2022, 11:10 PMgtx 1660 Ti 6GB here - u
user
02/07/2022, 11:20 PMbreAD - a
Awesome Face (GTX 1660 Ti)
02/07/2022, 11:28 PMNvidia Bread - c
Couch
02/07/2022, 11:29 PMtime to set up mmi - c
Couch
02/07/2022, 11:30 PM600 steps up to 0.24 val loss and i still get shitty attention - c
Couch
02/07/2022, 11:42 PMnever mind mmi dont have pre trained model - t
The Wiggles and High School D×D
02/08/2022, 1:11 AMI should really just try to develop my own voice models while I figure out how to finish building this machine - t
The Wiggles and High School D×D
02/08/2022, 1:12 AMBoth English and Japanese - t
The Wiggles and High School D×D
02/08/2022, 1:12 AMIn their respective programs to clone their voices and at least see what data I need - t
The Wiggles and High School D×D
02/08/2022, 1:12 AMBesides, I tried cloning an English voice before