I mean the alignment

I haven't saw yet.

this is one looks good

but yes. The aligment changes.

Any diagonal ones?

From Tomás Rebero.

uh not this one

I'm sorry. My audio doesn't hearing so good.

hmm yes the dataset is the one to blame here

So. This happened here.

Any changes for Sigma and Denoise Strenght?

in Tomás Rebero Dataset.

jesus

sigma should be 0.9 or 0.8

and denoise should be 0.0345

wowzers

why did you even change those lol

Tomás says...

I also have a TTS voice out of my friend from Iowa, but I don't want it on uberduck

@User I have a question. When the AI tries to synthesize does it read the whole text first and then synth or just synthesize while reading? Because sometimes I notice you can trigger a character saying a sentence in a certain way but not the other way around. Example: making Minnie say “we wouldn’t want Mickey to....” triggers intonation but not “...., now would we?” At the end of the sentence.

It reads the whole text first—a.k.a. encodes it into an intermediate latent state—but when generating audio, mostly pays attention to whatever it's generated in the previous spectogram frames

How to reduce the background noise of this audio?

It works

A sample of my Ashley TTS

Thanks!

"Bonjour papa Gainsbarre" (Hello Dad Gainsbarre)

But the TTS is English based, is it possible to generate a french one?

so what's a good aligment chart?

Using LALAL.AI obtains a vocal results?