Also 64 res is a great base resolution

jesus, can people stop making image algorithms

they're already super good lol

~~yeah like why dont they make themselves a bitch algorithm ~~

this looks promising

code is 5 years old tho

Lol when I first read this I thought it was tacotron without the need for transcripts RIP me.

uh...

getting out of mem errors on the arpabet notebook with a batch size of 2

404 wavs

are one of your wavs ridiculously long

oh wait that may be why

Yep, 28 seconds long

correction, 36 seconds

Eh, I'll cut them later it's just a test model anyway

yeah 12 is like the limit of what tacotron can handle

Hey guys, so I've been looking into Coqui TTS, and it seems really easy to get up and running. What's cool is they pretty much have most of the TTS models you could need, so you don't need to go and train a ton of stuff separately, and it's one consistent interface for everything. Has anyone ever experimented with this?

i think that's what uberduck is trying to replicate with pipeline actually

I can kind of see that, but the pipeline only contains Tacotron, which at this point is quite old. I wonder why they didn't just implement a version of Coqui that could theoretically be run on a server somewhere. It seems really easy to get up and running, although I haven't tried yet.

At this point I just wanna shove a bunch of wavs in a folder, transcribe like, 5 or something, and let the ai train and figure out the rest of the transcript itself as I use it.

I guess the more voices w/transcripts the more accurate it could be. Should think about the logistics of that more if anything

That's the dream.

I'm the weird guy who actually enjoys doing transcripts after sitting there for like 15 minutes lol, and I think having the automatic speech recognition take care of the bulk of it makes it easier, but yeah that would be amazing.

I think for the amount posted in #835647732453605376 , I think a rough dragh could be made

How do you spell that?

Draft?

Yeah

Damn ok, got Coqui installed on Mac OS with no issues at all. Got it to synthesize speech with Tacotron and going to try training next.

Lmao got a bit lost at the end there.