this is a great model...
just a kudos
not all models 'work', this one does.
I converted it to work with Auralis, and so it's both fast and it sounds like him
thx thx lol
which ones don't work?
Specifically any that you'd want to work?
I Realize now you probs meant 'work' as in 'sounds good' not 'broken' :p
the Morgan Freeman doesn't 'work' for me, as an example. It neither sounds enough like him, or it makes weird changes/sounds in places.
That said, I did open a bug with Auralis:
https://github.com/astramind-ai/Auralis/issues/35
because I'm not seeing any difference between using this model (converted) and the stock engine, with the same audio wav.
It would be nice to make a collection of 'good wavs' and figure out what needs fine tunes, and what does a good oneshot.
ohhhhhhh,
lol actually I just need to fix that model I think,
thats one of my old crappy fine-tunes when I wasn't very good at it
I'll see if I can fix it tonight and hit u up lol
Been trying to standardize by fine-tune repos by making the "ref.wav" file be what file should be used as the ref
Here I tried updating the morgan freeman model by re-training it, what do you think?
https://huggingface.co/drewThomasson/Morgan_freeman_xtts_model/tree/main/7_epoches
His voice seems to be quite a challenge nlg, there isn't very much good training data of him out there