and when there's not enough to fill 1 batch it just does 0 batches per epoch

also while talknet can't stroke it's really not gonna work to any reasonable degree with 4 words

i say wait until we get a massive multispeaker model and try to ask it to emulate ben's voice, like real-time-voice-cloning

no

😭

i hope this will go well

during the training what is the ideal loss? or recommended epoch?

is this bad?

it means that the test audio it generated was a bit too loud, i think it's normal for that to happen while it's training

Grimlock at Epoch 81

I weeded out a bunch of bad audio.

(→_→) Why do i keep getting a list out of range error on these new training colabs, ik the list matches the waves exactly wtf

either you have an extra

|

somewhere or there's one missing or maybe an empty line (e.g. at the end)

It was actually only because my phone was a weird width with text wrapping that a line got seperate into 2

`` file.wav|lorem Ipsum ```

Like that, but if the text was longer

I made a Needle Virus but I typed "NEEEEEEEEEEEEEEEEEEEEEEEEEEDDDDDDDDDDDDDDLLLLLLLLLLLLLLLLLLLLLLLLL l VVVVVVVVVIIIIIIIIIIIIIIIIIIIUUUUUUUUUUUURRRRRRSSSSSSSSSES" because My Needle Model got Failed so I need make a Second Version of Needle Model and I used LALAL.ai for Removed the Background

And the Model had 36 lines

oh yeah how do i do multimodel exactly

id like to revisit the megamind dataset i made that somehow isnt enough for the ai

even though its more than 15 minutes

hmm

definitely something wrong

how so?

like, are we not supposed to use the default training one? I'm trying to use it and it just spits out an error during the data check

the file is right freaking there

how do i use these?

Getting annoying RuntimeError on latest synthesizer via pipeline. First time using new notebook. Tried both with GPU and without. Any ideas what could be causing it? Making a Glados clone: 1-5_O0VV9M-FirMlkHYi4U0k9mjqGPElF

the problem is from data.pkl file