# machine-learning
- s
SuperLuig
06/03/2021, 2:18 PMi think i just got a bad model while training it lol - s
SuperLuig
06/03/2021, 2:18 PMbut you can hear bits of his voice - p
pi
06/03/2021, 2:19 PMyou're using "sounds good" very very liberally - s
SuperLuig
06/03/2021, 2:20 PMtrue - s
SuperLuig
06/03/2021, 2:20 PMxd - u
(Edd) bruh moment
06/03/2021, 2:23 PMGuys - s
SuperLuig
06/03/2021, 2:23 PMidk this model has proven an absolute pain in the ass to train - s
SuperLuig
06/03/2021, 2:23 PMhalf of the time the chart comes out disfigured beyond belief - u
(Edd) bruh moment
06/03/2021, 2:24 PMI one-time tried to train my dataset but it failed - s
SuperLuig
06/03/2021, 2:27 PMfuck it if anybody wants to examine it - p
pi
06/03/2021, 2:29 PMtoo much background audio here for example - p
pi
06/03/2021, 2:30 PMalso it's a very emotive voice - p
pi
06/03/2021, 2:30 PMso you'll see improvements from increasing the dataset size - p
pi
06/03/2021, 2:30 PMor only using clips where the character has a similar prososdy - u
user
06/03/2021, 2:31 PMtrained with background noise? - p
pi
06/03/2021, 2:32 PMwhat? - u
user
06/03/2021, 2:33 PMdid you train a model that has a background noise - u
user
06/03/2021, 2:34 PMtraining michael rosen - u
user
06/03/2021, 2:40 PMdoes training more improve the model? - p
pi
06/03/2021, 2:44 PMyour model sounds already quite overfit - p
pi
06/03/2021, 2:44 PMand your input audio seems to be quite noisy too - u
user
06/03/2021, 2:46 PMits from waveglow - p
pi
06/03/2021, 2:48 PMthat's not the culprit here - u
user
06/03/2021, 3:24 PMthis is not orginal source of michael rosen video - u
user
06/03/2021, 3:46 PMHerunterladen - w
Waltman13 (semi-active)
06/03/2021, 4:19 PMHe's comin out pretty good so far (trained with 44 samples btw, final model will most likely have 60 samples) - u
user
06/03/2021, 5:01 PMyou can remove background noise with audacity - k
kaijac
06/03/2021, 5:03 PMWho wants to make me a 24KGoldn voice lol - k
kaijac
06/03/2021, 5:03 PMOr Ian Dior