Training settings
Hi, I tried your rvc model, it works really good.
Wondering if you can share your training settings, tips or dataset(if can)
I tried training myself but not as good as your Nanami model
Thanks
oh hey thanks
this was actually the very first model I trained so I wasn't really familiar with all the settings. I believe I just trained it on crepe for 200 epochs without even checking the loss graph lol.
It could be bcz the dataset I put together is really clean; it had no bg noise or any other voices and it was of good quality. The dataset is always the most crucial part of the process imo
Really glad you like my model and I hope that helps :D
Hi, thanks for replying
Just wondering about "no bg noise", are you just manually filter out audio with bg noise from dataset or you use some tools to remove the bg noise
Thanks
It's a bit of both actually, tools like uvr or other vocal separators do most of the job for you but sometimes it doesn't get rid of all foreign noises. In that case I'd just discard those parts if I can't remove those noise without editing the audio too much. My rule of thumb is quality >>> quantity, so just find the best audio you can use even if you end up with a shorter dataset.
Good luck
Understood thanks :)