Singing sample:

make him sing this

Video removed in response to copyright claim.

Talking sample:

anyone know if its worth training to 600 epochs

or is it already good as it is

(0:15 skip) i think it sounds pretty good as is. a little bit of extra training might not be bad, though. go to 500 and make it save a model every 50 epochs so that you have working backups incase you overtrain it

also here's how it sounds alongside a track aswell, instead of solo i think its pretty capable as-is

aight will do after obamba model

v2 is very capable at 300 epochs tho, very nice

Joe Biden (RVC V2 Crepe) (500 Epochs)

alright ive done did it

500 epoch now

not much difference in quality tho, i suspect overtraining point is pretty close

are you using a colab/do you have a tensorboard to view data?

yup

its not started to go up yet but its diminishing returns

500 is more than enough, especially if you cant see much of a difference in quality

i will be checking out the 500 epoch version now, thank you for uploading

yup np

ill stop working on this one, at least till rvc v3 comes out

haha fair ^^

obama model:

Joe Biden (RVC V2) 500 Epoch

what if ballin but country???

Video removed in response to copyright claim.

I hate the picture. Poor Rin...

Made him sing Not Like Us

What the sigma

i want him /j

Rjhhhgjgh

Great model, used it here: youtu.be/ASthiLrGmfs

MINI SODA!!!