# machine-learning
- u
(Dawn) Will Draw Fictional Women
12/11/2021, 12:03 AMi wouldn’t even worry about it then - u
(Dawn) Will Draw Fictional Women
12/11/2021, 12:03 AMgenerating without reference audio isnt even an option on the site - u
(Dawn) Will Draw Fictional Women
12/11/2021, 12:04 AMit’s probably because the pitch and duration predictors are typically not even cared about when making talknets cause its only uses for reference audio - b
beepbopbus
12/11/2021, 12:04 AMi see - b
beepbopbus
12/11/2021, 12:05 AMthx for clarifying - s
SpamZow (WoomyPearl)
12/11/2021, 2:01 AMMario (Latin American Spanish) (By WoomyPearl) - s
SpamZow (WoomyPearl)
12/11/2021, 2:01 AMAlso, I want to make a model out of this. - s
SpamZow (WoomyPearl)
12/11/2021, 2:01 AMBut I have a few questions. - s
SpamZow (WoomyPearl)
12/11/2021, 2:02 AMMario says “Game Boy” in this ad, which is an English phrase. - s
SpamZow (WoomyPearl)
12/11/2021, 2:02 AMAnd he says “Va benne,” which is an Italian phrase. - s
SpamZow (WoomyPearl)
12/11/2021, 2:02 AMDoes it ruin a model that is meant to be in Spanish? - s
SpamZow (WoomyPearl)
12/11/2021, 2:03 AMAlso, I remember that Mario said Fantástico in Mario Kart Wii, so there’s another WAV for the dataset. - s
SpamZow (WoomyPearl)
12/11/2021, 2:04 AMGonna find more Spanish words Mario said. - s
SpamZow (WoomyPearl)
12/11/2021, 2:05 AMI predict that there will be like 10 to 15 WAVs total in a dataset for something like this. - s
SpamZow (WoomyPearl)
12/11/2021, 2:06 AMBut then again, UberDuck has characters with even less words such as FNF Girlfriend. - s
SpamZow (WoomyPearl)
12/11/2021, 2:08 AMMaybe I can even do what was done for the Hitler model and partially bruteforce Mario by giving him English phrases transcribed as Spanish phrases so that he could sound a bit better. - s
SpamZow (WoomyPearl)
12/11/2021, 2:12 AMOr I’ll maybe train another Mario voice (from a TV show or something) rather than Charles Martinet’s Mario. - u
(Dawn) Will Draw Fictional Women
12/11/2021, 6:19 PMbruh - c
Couch
12/11/2021, 6:25 PMget rid of the .zip extension - u
(Dawn) Will Draw Fictional Women
12/11/2021, 6:28 PM😳 - u
(Dawn) Will Draw Fictional Women
12/11/2021, 6:29 PMwill it still work cause i got the hifigan from justinjohn - m
matty1406
12/11/2021, 10:12 PMstonks??? question mark? - m
matty1406
12/11/2021, 10:12 PMstill looks messy but its still very early - u
user
12/12/2021, 8:41 PMwhitey from eight crazy nights test. it's like a little over 5 minutes of data to train on but i know i can add more than that - j
JealousJulian03
12/13/2021, 4:00 AMA failed tacotron synthesis notebook test of my model of Katsuma from Moshi Monsters: The Movie. He’s supposed to be saying “This is a fake message.” But it comes out as something untranscribable. Would apprechiate if someone could tell me what’s wrong. - p
Pekka
12/13/2021, 7:05 PMIs there a program that can split an audio file into it's sentences automatically based on silence? I remember seeing it somewhere. - h
hecko
12/13/2021, 10:56 PMauditok i believe https://pypi.org/project/auditok/ - c
Couch
12/13/2021, 11:09 PMsox - u
user
12/13/2021, 11:52 PMHi, ehm i have a question: how long does it take for the page to uploud an artist's voice? (i was waiting a lot lmao)