Spaces:
Running
Ignores chosen voice
No matter which voice i choose in FIne Tune it uses the CLONED voice. The app starts with a cloned voice visible and will only use that. New browser, no cache, but cleared it anyway.
if you choose morgan/freeman fine tuned model and want morgan freeman as voice, so choose the morgan freeman in the voice list.
if you choose morgan/freeman fine tuned model and want morgan freeman as voice, so choose the morgan freeman in the voice list.
That is for VOICE CLONING which is different than a fine tuned model. VC is pretty much a waste of time.
fine tuned is a model, xtts always need a voice attached to the model. We gave the freedom to the user to play from any model to any voice.
btw I'm the developer of Ebook2Audiobook....
I can confirm this is true,
thx @zermok
It should be automatically selecting the matching voice to go with the Fine-tuned model selected....
If it does not then you'll have to manually make sure that the Fine-tuned xtts model and the VC voice selected are matching
why it should be? an xttts model can be used with any voice, even if it's fine tuned....
if I force the fine tuned to use the original voice, maybe the users won't understand it can be used by other voices...
fixed in the coming update
That's a good point... hm
sorry, notifications got turned off. Been using on and off since jan or feb. It used to be that when you selected a voice in fine tuned model that that was the voice used regardless of what was showing in the clone voice section. What I meant was that nothing selected in Fine Tune models was used. It was always using clone voice. Glad it's getting fixed. Thanks
yeah were still working on fixing it, right now E2A current build is broken
so the pre-build E2A Docker image and Hugginface space are stuck at a The Last working Build until we get it Fixed :/
Thank you for your patience,