Just uploading a few of these incase i accidentally end up deleting the sub-pretrains and i dont wanna retrain it all over again. I still do be lazy uploading the rest though.
pretrain info
7subpretrain = 40K pretrain. trained on 1 hour of singing studio quality japanese female vocals for i forgot epochs lmao. I usually use it on japanese female datasets that cant hit notes for shit or less than 15 minute datasets. You can also use it if your a lazy ass who cant be bothered to train for more than 50 epochs
GenshinENPretrain32k = 32k pretrain. trained on various english genshin voicelines. Usually used for EN Voicelines or well english speaking characters that are below 10 minutes. I trained it 32k cause its less intensive for realtime even though i dont voice chat. Yes i know a chunk of them dont want to be used for A.I and I simply do not give a shit. Sue me
how to use or whatever
- just copy the path of both d and g pth
- paste it here. remember to put them correctly (it says d and g path)
- would look like this
- now just train normally i guess lol
how you make the sub pretrain tho? you just train it on top of the existing pretrain and the d and g pth file it generates is the sub pretrain. if you have atleast 20 or around 100 hours of audio please just make a full on pretrain and not a sub pretrain cause it would be overfitted as hell lol