That's cool

I need to retrain my german model because it corrupted and i want to cry about it I'm so scared i did something wrong and that it won't work again though

my my does it sound very promising now

before training hifigan for comparison

So happy with the results

I think that's enough examples for now

Nice nice

is there still a notebook for testing crust models before putting them in uberduck?

Yeah, the pipeline tt2 one in #841437191073955920

testing Steve1989

Hey man could I get a link to the HIFIGAN notebook you mention? It's been rm'ed from #841437191073955920 for some reason

What does /r/ sound like in Polish?

Tapped or rolled

Just like in Spanish

hey zwf, do you mind sharing with me how you created tts .wavs with musical rhythm to them using radtts? I keep trying to use this notebook from #841437191073955920 to synthesize a rapping Lupe Fiasco voice but can't: https://colab.research.google.com/drive/1SKu2xRJy5q1wzuP5CSO8dJ-Nf-UIKz0K My process is to use the Tacotron2 training notebook in #841437191073955920 to train a tacotron model using .wavs of Lupe's voice, then to finetune that tacotron model using that notebook I just linked to above, and to then pass the outputted model and .json from that finetuning notebook as vocoder inputs to the radtts inference.py command. Is that anywhere close man?

hey guys, was able to get access to 4x80gig a100s for a while, is there a repo that easy to setup to fine-tune VITS or GradTTS models?

For vits you can try espnet2

There's also coqui https://github.com/coqui-ai/TTS

For gradtts i think i used the original repo

you able to train that?>

i havent succeeded yet i tried once then got bored. rn im focusing on other things though

is there any documentation to train it?

yeah, on the repo there's a link to the docs

What am I doing wrong?

Oh nvm I'm being stupid lol

man idk what happened

training looked so nice the first time

how do i train tortoise tts?