I’m looking to train a singing voice model entirely from scratch (no pre-trained weights) I have sourced around an hour right now. Most RVC/Applio setups seem to require a pretrain. Does anyone here know of an app, tool, or workflow (maybe even a Kaggle notebook) that lets me train a high-quality singing voice model purely from my own data? Any tips or links would be awesome!
#Need help in training without the pretrain
clever mesa
thorny mountain
I’m looking to train a singing voice model entirely from scratch (no pre-trained...
What's your PC GPU and OS?
also be aware that training without a pretrain is really not recommended, and won't help, especially with just 1 hour of training
the pretrain is used as a base for training models infact
clever mesa
the pretrain is used as a base for training models infact
yeah but since i am training arabic singers, it sounds bad
its as if an american trying to immitate an arab
clever mesa
also be aware that training without a pretrain is really not recommended, and wo...
yeah i know that without a pretrain a minimum of 24 hours of data is required
i am ready to source that
clever mesa
What's your PC GPU and OS?
i will be buying a new laptop, prolly will be a 4090 mobile with i9 14th gen, and obviously windows as the OS.
thorny mountain
i will be buying a new laptop, prolly will be a 4090 mobile with i9 14th gen, an...
in Applio -> Training tab -> Training section -> Advanced Settings -> Uncheck Pre-trained, and it won't utilize pretrained models
you can use this on any Applio, local, kaggle, lightning.ai or Colab
You can check also https://docs.aihub.gg for more info :)