-
|
I have studio data (clean audio + text) per speaker. Which model is the best? |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 2 replies
-
|
Coqui includes a large number of pretrained models that you can evaluate for your use case, e.g. compare the different English models trained on LJSpeech. Note that "best" can mean many different things (naturalness, intelligibility, speed, efficiency, ...) |
Beta Was this translation helpful? Give feedback.
-
|
Dear @eginhard , Do we have any model that combines naturalness, speed, and efficiency? If yes, could you please recommend it? Additionally, do we have any model that supports emotion recognition? Is there an "all-in-one" model available as well? Thank you. |
Beta Was this translation helpful? Give feedback.
-
|
In my opinion and experiments, vits, and xtts_v2 is the best models. Xtts_v2 much more difficult to fine tune. |
Beta Was this translation helpful? Give feedback.
Coqui includes a large number of pretrained models that you can evaluate for your use case, e.g. compare the different English models trained on LJSpeech. Note that "best" can mean many different things (naturalness, intelligibility, speed, efficiency, ...)