Fine tuning XTTS-v2 on new language
#64
by
AllanK24
- opened
My intend is to fine tune the model on new language, which is Turkmen. Turkmen language is a language of Turkic languages family that includes also: Turkish, Uzbek, Kazakh and other languages. Since XTTS-V2 is already familiar with Turkish language, I was thinking of fine-tuning specifically Turkish XTTS-V2 on Turkmen language. Is this a right approach and is it achievable to get a good performance if my team does so? Can the model adjust it's Turkish knowledge to Turkmen language with fine-tuning?
Would be glad if anyone knows anything about it!
I don't think there will be any problem.
Any updates on your experiment? Did perform well on the Turkmen data?