# machine-learning
- b
BrokeSkill
10/15/2022, 9:09 PMSo cornell movie Dialogs? - h
hecko
10/15/2022, 9:16 PMi guess - c
Cris140
10/15/2022, 9:18 PMI don't know which gpu to get, 3060 ti or 3060. I want it to train models locally, the TI version has more speed but the normal version has more vram, which one do you guys recommend? - l
Lexi (delulu posts on the daily)
10/15/2022, 9:19 PMwhy not save a fuck ton of money and get both - l
Lexi (delulu posts on the daily)
10/15/2022, 9:19 PM2 gpu is better than one - c
Cris140
10/15/2022, 9:19 PMIt will be hard as hell to get one, imagine two XD - c
Cris140
10/15/2022, 9:19 PMThey're expensive where I live - l
Lexi (delulu posts on the daily)
10/15/2022, 9:20 PMah ok - l
Lexi (delulu posts on the daily)
10/15/2022, 9:20 PMhmm - b
BrokeSkill
10/15/2022, 9:28 PMSell ur liver - l
Lexi (delulu posts on the daily)
10/15/2022, 9:32 PMyou could sell part of it it can grow back - l
Lexi (delulu posts on the daily)
10/15/2022, 9:33 PMorgan selling is super risky - l
lljwsyn
10/15/2022, 11:13 PManyone know how this is created? - l
lljwsyn
10/15/2022, 11:13 PM - l
lljwsyn
10/15/2022, 11:14 PMif you add instrumentals it to make it sound like speech - l
lljwsyn
10/15/2022, 11:14 PMill give an example - l
lljwsyn
10/15/2022, 11:15 PMit like clones it with speech - l
lljwsyn
10/15/2022, 11:16 PMit might be hifi gan 2 - l
lljwsyn
10/15/2022, 11:16 PM - l
lljwsyn
10/15/2022, 11:16 PMi havent seen an implementation yet... - l
lljwsyn
10/15/2022, 11:17 PManyone know if there is one? - l
lljwsyn
10/15/2022, 11:18 PM - l
lljwsyn
10/15/2022, 11:18 PMthis might be something like it - l
lljwsyn
10/15/2022, 11:18 PM - l
lljwsyn
10/15/2022, 11:19 PMit is generative enhancement - l
lljwsyn
10/15/2022, 11:19 PMill check it out - l
lljwsyn
10/15/2022, 11:20 PMthis should be added to uberduck tbh - l
lljwsyn
10/15/2022, 11:21 PMill put it in #835971020569051216 - h
hecko
10/15/2022, 11:49 PMi see you've fallen victim to the hifi-gan confusion - h
hecko
10/15/2022, 11:50 PMthere are two entirely separate projects named hifi-gan one of them removes echo and unwanted noises, the other turns spectrograms into audio (and we use the latter)