Seems fine : Abonnez vous a mon Discord, je donne des Cauet Burger. | Subscribe to my Discord, I give some Cauet Burgers.

nice

I can try to train it with ultimate voice but there some echoes

well, if I'm not lazy, I'm going to do the voice of tommy vercetti in tts format

the first model i made was extremely overfit, so i'm redoing it. i can't tell if it sounds good enough yet or not, so i'm posting a sample here. how is it? good, bad?

additional info: 234 samples totaling at 10 minutes and 35 seconds, 108 epochs, and a validation loss of 0.131064. i used the hifi-gan notebook with tacotron_override for the above sample. also, here's the chart from the training notebook:

sounds ok

it's probably not overfit

I assume you don't understand what overfitting means

in this case, it means the model was trained for too long and replicates the dataset too closely i think? i'm probably wrong, but that's how i understood it when i looked it up lol

the reproducing the dataset is not the primary problem when overfitting tacotron

but your voice models is ok

so don't worry

if you want to make it sound better you'll need to collect more audio files

this piece of advice does not bode well for the other sc6 models i plan to make :'D

also thanks for the feedback! i appreciate it

Another French model test (Worst due to a lack of wavs) Tetris pas beau | Tetris not beautiful

how do i train my dataset?

many voice lines are missing, they are 256 don't ask me why I'm on another computer

= 0

👌

I missing 74 left to finish

Training the Rigby model RN

nice

I'm done with tommy vercetti voice

Bruh if someone can get me a battle droid voice from star wars I will do the naughty things adults do for pleasure to you.

I would also poop my pants a little

@User Do you make voices?