and you miiiight want to trim silence from the end of said training data but idk how much that does

there's nothing on stackoverflow

it's a test voice with 30 sample files

nevermind

fixed it

and now it's training

progress

30 samples

The graph doesn't look too good

nothing much has changed

this is only a test voice

the samples

well

idk

any better or worse

getting better but still looks poor

yeah I had to stop training because > well dad wants us to watch a movie and he doesn't allow me to use my phone or computer during the movie > and he times it so that it's close to or exactly bedtime when it's finished > and tomorrow's sunday so grandma has an excuse to put us to bed at 8:30 or 9 since we have church tomorrow

needless to say the voice produces no audio

this is the final product

yos i need some help

i need to convert numbers into ARPAbet

e.g 1 -> one ->

W AH N

or 22 -> twenty two ->

T W EH N T IY . T UW .

I mean, to me that looks correct? Not sure which numbers you need?

i need to create a python function to automatically extract the numbers from a string into their word equivalent

The man went into store 22 for questioning.

->

The man went into store twenty-two for questioning.

then i can just run that whole sentence through arpabet

theres probably an nlp library that does exactly that

theres also the english text cleaner from tacotron2 which normalizes numbers into words

oh cool