base_model

select

Oh now I see it. Thanks hecko.

Nevermind, wtf is this?

What's the issue?

WHAT THE FUCK!!!

WHY AM I GETTING A RUNTIME ERROR EVERYTIME?!?

this really is support 2 💀

Dawn do you know what the issue is?

I think I figured it out.

There we go! Now it's training!

This channel isn't meant for support. This channel is to send tips and tricks to make training a model easier

Oh shit, sorry Pikachu. Next time I’ll use the support channel. Didn’t know man.

Hii

new channel moment

When making a tacotron model with more than an hour of audio, when the loss plateaus, stop the training and re run the training with learning rate of 0.0005

have about 350 audio samples at best

0.00*0*5

three zeroes

this is important because with two zeroes you're increasing it fivefold instead of decreasing twofold

What’s the best website to create ai on mobile for uberduck i’m trying to make a Scarlxrd ai Well two of them actually

1: would ask in support next time, 2:if there's no pre separated audio there's sites like mvsep.com or lalal.ai 3: #841437191073955920 for training

Thanks I actually made a drawing of what I might use for the Scarlxrd AI

i'm trying to make talknet el chavo what do i do on the other 2 links?

Also another thing how do I blend acapellas

Here's a tip, plush encounter. Your dataset should be a zip file containing audio files not in any folders in the ZIP file, just in the ZIP file, because it's faster. The train and val filelists should be something like this 1.wav|Your text here. 2.wav|Your text here. 3.wav|Your text here. And so on. output_dir should be something like /content/drive/MyDrive/talknet/el chavo.

i know that

but not for talknet

When training the Mario Teaches Typing TalkNet 2 model (which I probably need to retrain), I used the same filelist for both train and val.

Here's an example of a ZIP file for TalkNet 2/what I'm talking about: