The Uberduck one

the cell next to the mel cell on the picture doesnt look like the uberduck training one, i think...

are u sure thats the uberduck one

@User yes, its just the check for missing files

@UserI just tried to use the unzip, but:

@matty1406 can you send me phone guy's dataset so i can try him on talknet?

Hey, still new to this, but I have 4 minutes of isolated vocals that I want to use with TalkNet. Not sure of the process here, specifically what the train_filelist.txt and val_filelist.txt are. Using the TalkNet training colab in the #841437191073955920 channel, how would I go about training a model on the vocal samples I have?

awwww my first training attempt

nice lol, promising

make it say the number

Hi there Matty, what solution you had for that issue? I am now facing the same problem on my model training

It meant one of my audio files was a stereo and not mono

I Was Saying: Do PinkFong In Uberduck Voice

Oh thanks, thats off I ran it all through the checker, well I guess I should double check it then

iirc it doesnt check if its 22khz mono

I resolved that issue, thank you alot. Right now I am looking why isnt it capable of seeing an file which is clearly there, I will double check everything for possible mistkaes I could have made first before bothering anyone here

Oh I think I got it, I need to add the containing folder filepath of the .npy files into the list.txt instead of just filenames

Thanks for everything, running like a charm now

Now I just need to wait a couple of hours

hm here's a weird thing i have 45 wav files, and with a batch size of 14 the progress bar was 3/3 so i thought it was ignoring 3 files, so i set batch size to 15 and the progress bar is 2/2??

and i can't find anything obviously wrong in the notebook so it might be an issue with the tacotron code itself

around 100 epochs, this is so cool, didnt even take that long

"from uhhh sweden"

it got the voice down even in replicating the intro

awesome!

yeah im very surprised

for my first ever try? not bad

do i train it more? can i continue training it just from the same notebook (i terminated it because i had to test the model)

How many files you had for this?

wavs I mean