Spaces:

parler-tts

/

parler_tts

like 732

Running on Zero

Using identical prompt conditioning the quality of audio produced is not consistent between web app and downloaded version. I am using CPU and speed is still impressive but samples frequently sound heavily distorted, over emphasise certain syllables ("H" in particular).

Is is possible to provide more information on how the web application is parameterised? I have tried both mini & large models.

Thanks,
Liam Barry
(Arm employee)