ok so like i m trying to train an ai model but every time i have (using applio with rvmpe) it just leads to the audio sounding like this; the dataset is 20 minutes (chopped into 3 second sections) of speaking at the recommended volume so i am not sure what i am doing wrong. i have let this go from as low as about 10-20 epochs to as high as 500 epochs and it sounds roughly the same. if it makes a difference i untick "pretrained"
#training model doesnt work (output is random noise)
faint mango
ok so like i m trying to train an ai model but every time i have (using applio w...
trying to train from scratch ("unticking pretrained") will always sound bad
hard sinew
oops
hard sinew
trying to train from scratch ("unticking pretrained") will always sound bad
does training with a pretained voice like affect how the character sounds in any way? i guess i unticked it cuz i was a lil paranoid about ruining the "integrity" of my characters voice
faint mango
does training with a pretained voice like affect how the character sounds in any...
your worry doesn't make sense, again haven't you known without such building blocks it will always sound bad?
hard sinew
no i have never done this b4
faint mango
<:nails:1159569314848972891> no i have never done this b4
no it's just your misunderstanding, the original rvc by default uses its default pretrain for corresponding sample rate (f032kD, f032kG, etc)
and so in applio the "pretrain" is checked by default to actually use it
likewise, you can't build a program without using a compiler, even building from some assembly code still needs a compiler
hard sinew
no it's just your misunderstanding, the original rvc by default uses its default...
ohh
hard sinew
likewise, you can't build a program without using a compiler, even building from...
ok yeah i gotchu now
so basically i went "oh wow look at all this x language on the floor but why wont the computer just work it!!" forgot the compiler 😔
tranquil reef