disappointing. and I live in th phili pines

Now i can't make any more voices

Nah you will, just wait about 24 hours

I'm trying to train a Pipeline version of Optimus Prime.

I found the problem. There was a typo.

Can I use the Uberduck Pipeline Training notebook without the GPU

training no

https://drive.google.com/file/d/1-35h6Uw2XOT0ewy2gJP11sCfl9dZZw9l/view?usp=share_link Is this okay, or do I have to train it longer?

It's at 20th epoch

This is the dr. crygor smooth moves model, original VA Tim O'Leary

I don't understand what he says in this audio clip

yes its muffled, needs further training "if it's going to be next to something else, I think it'd be best to have extra space on all sides."

The 40th epoch result (https://drive.google.com/file/d/1-3z5S2FYWIDNUhAuldq4vZVEkNjqM-Bq/view?usp=share_link)

question, is there a program to auto detect voices and chop up .wav datasets automatically? or some pre-existing code that is for the automation of creating chopped up audio files? like, if it's an interview or random features, the program can autodetect what voice it is and i can just throw out the other voices if they need it chopped every 5-10 seconds? like throw out 'um', weird stammering, and white space and trim the audio automatically and label speaker 1 from speaker 2, etc

Remember my Crygor WWSM model? It's now at 120th epoch

180th epoch

I think it needs to be treated with hifigan

Is he having a stroke?

I did post a couple 80-epoch clips of my in-progress Pipeline Optimus Prime model in <#820353681567907901>

At 320th epoch.

Why is the loss graph perpetually going down slowly?

Might be time to put a enhancement with hifigan vocoder trainer

It's not flattening for ages! It's been 6 hours!

Apparently, the loss should dip below 0.1 though. I guess that's a big ask for voices that typically have metallic filters.

Like this? This is the latest test of my model.

like as in, "as in"

I can make out *some* of what he's saying, but not all of it. Is he supposed to be doing an anti-tobacco PSA?

It's in Japanese. I can't read it.