AlignMixin#

class montreal_forced_aligner.alignment.mixins.AlignMixin(transition_scale=1.0, acoustic_scale=0.1, self_loop_scale=0.1, boost_silence=1.0, beam=10, retry_beam=40, fine_tune=False, phone_confidence=False, use_phone_model=False, **kwargs)[source]#

Bases: DictionaryMixin

Configuration object for alignment

Parameters:

transition_scale (float) – Transition scale, defaults to 1.0
acoustic_scale (float) – Acoustic scale, defaults to 0.1
self_loop_scale (float) – Self-loop scale, defaults to 0.1
boost_silence (float) – Factor to boost silence probabilities, 1.0 is no boost or reduction
beam (int) – Size of the beam to use in decoding, defaults to 10
retry_beam (int) – Size of the beam to use in decoding if it fails with the initial beam width, defaults to 40