Training or Fine Tuning a Hindi Language VITS TTS Voice Model with Coqui TTS on Google Colab
I received a few questions about training a Hindi language VITS model. I couldn't answer them well enough, since I hadn't done it, so this is my attempt. If you have a decent quality dataset, and double check the character set, you should be able to make it work. However, if you have the ability, it is probably best to run Coqui at home. Colab is VERY slow for file transfers, and this will slow down your dataset loading time, and eat into your computing credits very quickly.
Colab Notebook:
https://colab.research.google.com/drive/1lKsO-_5TJE8q_Ot1-t5380hobxNL_SWa?usp=sharing
Other voice videos:
https://www.youtube.com/playlist?list=PLn3gJVMORUsTiGp5Jc37XID9jS9SHKUV1
https://www.openslr.org/103/
Multilingual and code-switching ASR Challenge Dataset - sub-task1
Hindi_train.tar.gz [4.4G]
https://commonvoice.mozilla.org/en/datasets
Select Hindi
Common Voice Corpus 12.0