recommit voices

remove voices..

commit voices and examples

Merge remote-tracking branch 'github/main'

Update readme with prompt engineering

Enable redaction by default

change quality presets

Use librosa for loading mp3s

Remove entmax dep

Fix default output path

k I think this works..

fix paths

getting ready for 2.1 release

misc fixes

Fix bug with k>1

more fixes

fix warning

Support totally random voices (and make fixes to previous changes)

move data

Add redaction support

Better error messages when inputs are out of bounds.

add support for specifying the model_dir

Add support for extracting and feeding conditioning latents directly into the model

more cleanup

Move everything into the tortoise/ subdirectory