Post by someone who works at Hugging Face. Parler is a series of two TTS models, which was trained on open speech, and has efficient generation. This was categorized as pio because the OP wrote “can’t wait to see what y’all would build with this!“.
Commenters ask a couple things:
- How does generation speed compare to other TTS engines?
- OP does not have hard comparisons
- Is it compatible with Apple Silicon?
- OP says yes, and that you have to pass in “mps” as the device.
- What is the list of 34 voice names?
- Another commenter answers this question.
- This is English only, right?
- A commenter answers this question as they have tried Spanish and the model failed to produce a coherent output.