That may very well be possible. I have not checked the data. I use data from Mark/Rampage. If you find anything, feel free to correct it and send it to me!
Do you have the latest version 1.3.0? Because it was a bug in an earlier version. Do you have a bin file in your main directory? Something like ggml-base?
Yes, sometimes it's not so easy to find the right words to make it sound natural Sometimes you have to do a little tricking because the intonation isn't perfect. But it's pretty damn cool (and also a bit scary) what's possible today Regarding your question with the models: Theoretically yes, it's all open source data that others have trained. You need good sound samples and you have to train the models using the files, then you can theoretically expand that. Maybe someone else can say more. But theoretically it is possible, yes.