# machine-learning
- m
mjh657
08/31/2021, 6:56 PMi am not getting any sound from the synthesis notebook - h
hecko
08/31/2021, 6:58 PMit doesn't work with k80, use this notebook which runs on cpu https://colab.research.google.com/drive/1NVA3ndxhYWsKn-zwh3NnzMMgoVdJ5xUx?usp=sharing - m
mjh657
08/31/2021, 7:01 PMthanks - m
mjh657
08/31/2021, 7:05 PM@User how do i use my own model - h
hecko
08/31/2021, 7:10 PMuncheck dictionary, paste model code in tacotron override, ignore the character dropdown, run all - u
user
08/31/2021, 7:22 PMArbiter end result - t
The Wiggles and High School D×D
08/31/2021, 8:13 PMHm Went back to take another look at my archives I could have a way to extract all the English voices to help me work on this further, though I have no idea if I'll have enough for it to be viable - t
The Wiggles and High School D×D
08/31/2021, 8:32 PMHm The operation seems to function but my disk space is too high, about over halfway of 68 GB - t
The Wiggles and High School D×D
08/31/2021, 8:33 PMLet's see how much I can lower it and if that helps - t
The Wiggles and High School D×D
08/31/2021, 8:36 PMNot sure if that'll do much for me, unfortunately... - t
The Wiggles and High School D×D
08/31/2021, 8:36 PMYou sure any of you aren't willing to take a crack at this? - t
The Wiggles and High School D×D
08/31/2021, 8:38 PMHosting the model for anyone to use is probably better equipped to someone else anyway - t
The Wiggles and High School D×D
08/31/2021, 10:01 PMYeah, I might be at a dead end Unless anyone can give me ideas for what else I ought to do (like a less intensive but equally spacious English voice collection), I'm stuck - h
hecko
08/31/2021, 10:02 PMless intensive but equally spacious? not sure what that means - h
hecko
08/31/2021, 10:02 PMdo you mean same amount of voices but less data per? - h
hecko
08/31/2021, 10:03 PMi'd go with mlp data even tbh - t
The Wiggles and High School D×D
08/31/2021, 10:31 PMFor encoders and synthesisers work, that'll do? - t
The Wiggles and High School D×D
08/31/2021, 10:32 PMMaybe I could deal with thousands of hours of audio if they didn't come from 10GB files I needed to group together - t
The Wiggles and High School D×D
08/31/2021, 10:42 PMWhere can I find a dataset of as many of those voices as possible? - s
Scramble
09/01/2021, 1:48 AMmy dude really posted a full south park episode - u
user
09/01/2021, 1:50 AMgonna go my bus - m
mega b
09/01/2021, 2:53 AMdelete this fam - m
mega b
09/01/2021, 2:53 AMor else a ban - m
mega b
09/01/2021, 2:53 AMlol - l
Lizardizzle
09/01/2021, 3:54 AMIf I press the stop button on the training, edit the batch size, and start it again, will that create problems? - m
mega b
09/01/2021, 5:13 AMIt won't - m
mega b
09/01/2021, 5:14 AMjust be sure to resume training on the same dataset - l
Lizardizzle
09/01/2021, 6:36 AMThanks, that changed the time from 8 hours to around 2-3 - h
hecko
09/01/2021, 7:06 AMif you mean pony voices then https://drive.google.com/drive/folders/1OKm39u19bKMuaeR3diKeKVVvsXc1ZNYC note that they use a different format so you'll have to convert them (and remove noisy clips while at it) if in general then uhhhh idk, common voice has variety but it's also hella noisy - b
BIA
09/01/2021, 11:05 AMfor smaller datasets, is tacotron 2 a bad idea?