You do not have permission to delete messages in this group
Copy link
Report message
Show original message
Either email addresses are anonymous for this group or you need the view member email addresses permission to view the original message
to MFA Users
Hi,
I am observing that the phoneme alignments from mfa v1.0.1 are more accurate than v2.0 for LJSpeech and other datasets when aligning with the downloaded english model. Is there some setting that I am missing regarding v2.0?
Basically I generated alignments using both versions and trained a FastSpeech2 model and the inferences that were trained from alignments using mfa v1.0.1 were superior to v2.0.