# machine-learning
- s
Super LJ The Original
08/07/2021, 4:19 PMWhat did you use to generate those lyrics - m
mega b
08/07/2021, 4:20 PMGPT-J - m
Marsag
08/07/2021, 4:24 PMi used a site called https://6b.eleuther.ai/ - s
Super LJ The Original
08/07/2021, 4:26 PMThanks - s
Super LJ The Original
08/07/2021, 4:26 PMHave a great day - s
Super LJ The Original
08/07/2021, 4:26 PMYou are awesome - h
hecko
08/07/2021, 4:40 PMthat's actually closing down in some time supposedly, here's one that's less likely to do so and also is faster https://textsynth.org/ - u
user
08/07/2021, 5:43 PMis this good?: - t
TheLoudWhovian2002
08/07/2021, 5:45 PMLyrics from "Right There" (the song that plays in the closing credits of THE BEARD: Special Ooghdition by Random Encounters) but the rest is AI generated - u
user
08/07/2021, 5:45 PMusing inferkit? next time, put it on #841461499946860576 - h
hecko
08/07/2021, 5:46 PMusing this actually - h
hecko
08/07/2021, 5:47 PMit doesn't look very good, but maybe you just got a short test clip could you send another graph from a nearby epoch - u
user
08/07/2021, 5:48 PMthen how about this? - u
user
08/07/2021, 5:48 PMi had to say this but it isn't working! - s
Scramble
08/07/2021, 5:57 PMuhh - s
Scramble
08/07/2021, 5:57 PMthat's not good. - u
user
08/07/2021, 6:00 PMyeah, i know then is going down and down till the time when trying it, makes only noise. - u
user
08/07/2021, 6:01 PM@User look! it's betterizing! (that word doesn't exist) - u
user
08/07/2021, 6:07 PMCopy codeNameError Traceback (most recent call last) <ipython-input-3-bcc7dc251105> in <module>() 33 34 # Load Tacotron2 (run this cell every time you change the model) ---> 35 hparams.sampling_rate = 22050 # Don't change this 36 hparams.max_decoder_steps = 1000 # How long the audio will be before it cuts off (1000 is about 11 seconds) 37 hparams.gate_threshold = 0.1 # Model must be 90% sure the clip is over before ending generation (the higher this number is, the more likely that the AI will keep generating until it reaches the Max Decoder Steps) NameError: name 'hparams' is not defined - g
Gosmokeless28
08/08/2021, 9:40 AMWhat is the minimum amount of audio files that you can train a TalkNet model with? - s
SpamZow (WoomyPearl)
08/08/2021, 2:02 PMI remember doing a dataset with 19 wav files and it was trained as a Talknet model. - j
JacobLenstar
08/08/2021, 3:55 PM250 epoches on sarah from ed edd n eddy - j
JacobLenstar
08/08/2021, 4:20 PMremoved a few problematic lines, looking better - h
Hugh G. Rection
08/08/2021, 8:02 PMWhat the hell happened to Dixie Kong after I finished trained her?! - h
Hugh G. Rection
08/08/2021, 9:32 PMOkay, so I made version 2 of Dixie Kong's voice, but she still sounds bad. "This voice model still ||fucking|| sucks." - h
Hugh G. Rection
08/08/2021, 9:42 PMI'm rejecting my Dixie Kong voice model off the Uberduck website, because it sounds very horrible. Had to make version 2, with 52 voice clips and batch size, and 250 epochs. She sounds okay, but still horrible. - p
PUMPKINEATER
08/08/2021, 9:55 PMYes. - f
FluffyGuy
08/08/2021, 10:04 PMInteresting.