Prompts to included natural Pauses, Chuckles, breathing noises etc.
#6
by
jaydip-tss
- opened
How can you promp these models to make the audio sound more natural with some pauses or laughs etc?
Hey
@jaydip-tss
, ultimately, the model's capabilities are limited by what's on the dataset. Here the datasets we've trained on came from audiobooks, so there's no laugh or proper pause control.
However, in my experience, the use of dashes (-) causes the speaker to pause, so you can emulate pauses if you want!