# machine-learning
- o
OctolingVladTTS
05/13/2021, 4:15 PMWhat do you think of the update with 32 voice lines? the audio is from the navy seal copypasta - j
JacobLenstar
05/13/2021, 4:36 PMi'm doing scooby doo and instead of writing everything as ruh roh, rikes, rot rogs, etc i'm gonna write out what he's saying and see how it works - j
JacobLenstar
05/13/2021, 4:38 PMor is that a bad idea - s
SidPlays_144p
05/13/2021, 4:38 PMI think that's the way (honestly I've asked myself that) - k
kitsuner
05/13/2021, 5:03 PMI’d say it’s the smarted way because Lucy loud speaks has a heavy lisp but when I used her I didn’t have to change anything to get her talking properly - t
Toasty
05/13/2021, 5:53 PMHey Pi, a while ago, you said that the problem with small datasets is that overfitting can happen if I'm not careful - t
Toasty
05/13/2021, 5:53 PMWould a solution be making the dataset bigger? - m
mega b
05/13/2021, 5:54 PMyes\ - t
Toasty
05/13/2021, 5:56 PMWell time to make the Yahtzee Croshaw 2.0 Dataset bigger - t
Toasty
05/13/2021, 5:56 PMIs 2 hours a good goal? - m
mega b
05/13/2021, 5:56 PMyea - m
mega b
05/13/2021, 5:56 PMmy markiplier model used only 1 video to get the audio and it was 30 minutes - m
mega b
05/13/2021, 5:56 PMbut it was non stop speaking - m
Manduybandspn
05/13/2021, 6:35 PMWhat happened to the Fozzie model? - z
zwf
05/13/2021, 6:36 PMremoved for improvements at the request of the creator - j
JacobLenstar
05/13/2021, 6:43 PMyeah fozzie was completely unintelligible and incomplete - t
Toasty
05/13/2021, 6:47 PMWell, after an hour, I only added an extra five minutes of audio 😆 - t
Toasty
05/13/2021, 6:51 PMTime to train it - t
Toasty
05/13/2021, 6:53 PMOK here's a problem. - t
Toasty
05/13/2021, 6:54 PMZwf told me: > if the training loss is decreasing but the validation loss is not, that means overfitting - t
Toasty
05/13/2021, 6:54 PMthe problem is, my tacotron training notebook does not tell training loss - t
Toasty
05/13/2021, 6:54 PMonly validation loss - u
user
05/13/2021, 6:59 PMThat's what Kermit noticed here. - t
Toasty
05/13/2021, 7:02 PMOoh, I should test it with the uberduck tacotron training colab. - t
Toasty
05/13/2021, 7:51 PMAlright, at 25 epochs, gonna test it rq. - t
Toasty
05/13/2021, 7:55 PMYES - t
Toasty
05/13/2021, 7:56 PM(this is using a sentence from the transcript) - u
user
05/13/2021, 8:07 PMWait - u
user
05/13/2021, 8:07 PMWrong channel