File size: 129 Bytes
873d140
1
This is an early multispeaker model, trained pre-ContentVec fix, to test timbre leakage. Speakers are not easily distinguishable.