tortoise-tts
tortoise-tts copied to clipboard
neonbjb
unable to get voice without accent/removing accented training from dataset
one of my voices, no matter what i do, develops a british accent. can we get any more community support on this? found a couple posts about adding more clips but nothing really. have tried playing around with what i've found in the API but unfortunately nothing. i've tried changing preset from ultra fast to hq and no luck. any/all help is appreciated, i'm savvy enough to use CLI, github, and open source projects, but i'm not thoroughly familiar with neural networks or high complexity things like this.
lower the top_p to about 0.1 I put mine on 0.0 and temperature 2.0 and works flawlessly, for me that is.
I'm now running tortoise-tts-fast and while I don't imagine that would make a difference, I'd figure it's worth noting. Still no luck. Truthfully I don't know what temperature or top_p are alike, no documentation on what either mean outside of existence to "fine tune." EDIT: I'm wrong about that last part. Managed to find some more stuff in the API about them. A little confusing but I'm still tweaking until I can manage to get American women's voices to sound American.
I'm now running tortoise-tts-fast and while I don't imagine that would make a difference, I'd figure it's worth noting. Still no luck. Truthfully I don't know what temperature or top_p are alike, no documentation on what either mean outside of existence to "fine tune." EDIT: I'm wrong about that last part. Managed to find some more stuff in the API about them. A little confusing but I'm still tweaking until I can manage to get American women's voices to sound American.
where did you find can you link?
@kalioiczys Did you manage to find a setting that works? I'm having the same problem of American women coming out sounding British.
fwiw, I had the same issue, american speaker ends up with british accent.
I changed my samples from speaker's younger voice, to speaker's slightly older voice, and the accent went away. YMMV
I'm now running tortoise-tts-fast and while I don't imagine that would make a difference, I'd figure it's worth noting. Still no luck. Truthfully I don't know what temperature or top_p are alike, no documentation on what either mean outside of existence to "fine tune." EDIT: I'm wrong about that last part. Managed to find some more stuff in the API about them. A little confusing but I'm still tweaking until I can manage to get American women's voices to sound American.
Were you able to find a setting that worked?
I'm now running tortoise-tts-fast and while I don't imagine that would make a difference, I'd figure it's worth noting. Still no luck. Truthfully I don't know what temperature or top_p are alike, no documentation on what either mean outside of existence to "fine tune." EDIT: I'm wrong about that last part. Managed to find some more stuff in the API about them. A little confusing but I'm still tweaking until I can manage to get American women's voices to sound American.
Were you able to find a setting that worked?
you can these setting under api.py
