Closed Moon-sung-woo closed 1 year ago
You can try playing around with the command line options for --silence_probability
(I'm not 100% sure if you can affect silence probabilities of trained dictionaries, since there's a lot of per-pronunciation modeling that goes on) and --boost_silence=1.5
or something greater than 1.0.
Oh and also double check that you're on the latest version, there was a bug recently fixed made it seem like there were no silences with --cleanup_textgrids
, but --no_cleanup_textgrids
showed the silence (https://montreal-forced-aligner.readthedocs.io/en/latest/user_guide/workflows/alignment.html#cmdoption-mfa-align-textgrid_cleanup).
Hi! @mmcauliffe . Thank you for the comments. I'll try as you told me. Thank you!
Hello, first of all, thank you for making a great MFA.
I trained the MFA model for Korean and checked the results, but there is no silence with all spacing units.
Can we get MFA results including more silence?
What parameters should I modify if I can get it?
Thank you.