# machine-learning
- t
TwoTimmm
11/02/2021, 9:58 PMWill be really problematic (and I will do a lot of them too probably) - t
TwoTimmm
11/02/2021, 9:58 PMI also tried training jhin but the result was trash - h
hecko
11/02/2021, 9:59 PMmight get started with overwatch first because i have mostly-accurate subtitle files for it already - t
TwoTimmm
11/02/2021, 9:59 PMHe couldnt even say his basic lines, the "You lack imagination" would turn into a garbled mess and "Youuu.... Lllwack.. Imaginshswnwnwnnnnnnnnnn..." - t
TwoTimmm
11/02/2021, 9:59 PMI am surprised my camille model is really good because I wasnt even trying - t
TwoTimmm
11/02/2021, 10:00 PMI Just threw in some random 20 lines of her, transcribed them and put it in training mode - h
hecko
11/02/2021, 10:00 PMi mean hopefully with a multispeaker model it'd do better - h
hecko
11/02/2021, 10:00 PMbecause it wouldn't be hyperfocusing on a few dozen clips from one voice - h
hecko
11/02/2021, 10:01 PMso it wouldn't override knowledge of how to speak english with that of how to perfectly enunciate a particular clip - t
TwoTimmm
11/02/2021, 10:01 PMMy camille model can speak arpabet, can say almost everything properly and copies her original lines really great - t
TwoTimmm
11/02/2021, 10:02 PMThe arpabet part. I didnt even know she could do that until I quacked as her in #842056330835656745 - t
TwoTimmm
11/02/2021, 10:02 PMShe turned out good kinda simply on random - i
IBob012
11/02/2021, 10:10 PMyeah for some reason low dataset models can use arpabet, very weird but at the same time useful - h
hecko
11/02/2021, 11:36 PMooh interesting - h
hecko
11/02/2021, 11:38 PMmaybe i could try redoing the a hat in time cruise seals experiment where i turned their transcript `r`s into `w`s so it'd keep the `r`s from the base model spoiler alert: it didn't, and just skipped over the `r`s when synthesizing but maybe with less training it could work - u
user
11/03/2021, 12:20 AMwhat are the best deepfake apps? - g
Gosmokeless28
11/03/2021, 12:21 AMI only know of FakeYou.com - u
user
11/03/2021, 12:23 AMno like for making ones - j
Justin
11/03/2021, 2:56 AMDeepFaceLab, SimSwap and FaceSwap - p
Puterboy1
11/03/2021, 3:37 AMIs there any way I can upload this video to be used in a text to speech converter? - u
(Dawn) Will Draw Fictional Women
11/03/2021, 3:43 AMmake the dataset and train it - b
Balloonya from Pokémon
11/03/2021, 12:06 PMWhy resources channel? What do i do on it? - s
Scramble
11/03/2021, 12:28 PMthe resources channel is for the best places to find clean voice clips of a character (the only thing in there currently is sounds-resource which is great for game characters and characters that had games.) - g
Gamma Prime
11/03/2021, 11:39 PMsounds-resource would be great for me if there were more Transformers games covered. - g
Gamma Prime
11/03/2021, 11:39 PMRight now, they only have voice rips from the 2007 movie game. - o
Obi Wan Kenobi
11/04/2021, 11:06 PMI even linked the transformers game modding discord the voices rip from 2007 movie game so that Hover Ex unfinished level won't have a talking female impersonating optimus and megatron - o
Obi Wan Kenobi
11/04/2021, 11:06 PM(Some voices had peter and frank voicing optimus and megatron) - o
Obi Wan Kenobi
11/04/2021, 11:06 PM(some didn't) - o
Obi Wan Kenobi
11/04/2021, 11:07 PMThat means hover ex will be finished in the making of the extended version of transformers the game (Mod)