Hey staff/people who have successfully submitted models for the site, what's the turnaround time been like for submitting models lately? As in like. the amount of time between when you submitted your model and when it went up on the site I know all my talknet models are already public and usable on the notebook im just really excited tfigsjtioighieow

do you think it's good to submit?

if you think it sounds good and coherent enough its probably good for submission

thank you guys

especially @User who made half of the dataset

Some models of mine get approved almost instantly, while also having a few I submitted around a month ago still not up

i see. i put the link in #835647711305793551, i'll see where i'll go now.

Hey there, I'm new to ML, I've just read the announcement and I'd like to try and train a model for "Pitch and Rhythm Transfer" myself. Which model architecture is used for that? I think the TTS one used Tacotron or something.

The pitch transfer one uses Talknet!

yup

normal speech uses tacotron 2, pitch and rhythm transfer uses talknet.

oh interesting, so it will require a retraining on talknet right

yes

oops wrong channel lol

i have 100 samples, what should my learning rap an epocs be?

thanks!

Damn I can't wait for the Snow White voice samples to get processed soon

as I wanted her to sing silly stuff

Besides getting Trump to do much the same nonsense

anyone know?

im training a model, does it matter much if the outputs look like this

some are really good some arent

You need the alignment chart to show an unbroken line from the bottom left corner to the top right corner. That would mean your input texts will be fully converted into speech.

im currently using the uberduck training notebook

what would be the best settings/requirements to get the best results

bro, could you give me the dataset for stop posting about among us guy real quick

@User here, the transcription txt is in the zip

the whole dataset is literally the video and that's it lmao

Lol