Description
Dataset collected from family guy episodes with him talking and singing.
Comments
Singing sample (theres a bit of instrumental bleeding cbf to remove it)

Awesome! I've been waiting for a Stewie Griffin (Singing) RVC model like this

@Delik [CN/EN] Out of curiosity, did you include the vocals of Stewie's cover of Bryan Addams' song in the dataset?
which one
ty

Stewie Griffin's cover of "Everything I Do (I Do It For You)"
yeah
i have that included

Nice
its like 22 minutes talking 5 minutes singing
@[not] MJ (Brazilian Sr. Mod) do you think this model is undertrained

how would i know

you'd need to look at your graphs
hold up
i stopped at 28k
^

what about d total

and mel

and kl
g mel and g kl?

yes

looks good
if all those three decreases in a 1/x shape how tf g total has a turning point

yeah they're looking fine too, you should keep training
which graphs should i keep observing?
these four?

yeah those four
so if all those four start to bounce up do i stop training

wait for a while, if they dont come back down (especially mel, d and g) you should stop yeah

you're welcome :)
@[not] MJ (Brazilian Sr. Mod) 1300 epochs😭
electricity price boom boom

i mean kl has reached its limit

and i think d total might be increasing

so idk
its not
my dataset is too good 1300 epochs havent reached its potential
@[not] MJ (Brazilian Sr. Mod) is this an example of a dataset being too good lmao

damn

> maybe
you want samples?
@[not] MJ (Brazilian Sr. Mod)

dayum that sounds great
solved the problem using high school physics knowledge haha

Stewie.
Add a comment
Samples
Pitch