2.1 Changelog

2.1.6

• Fix for issue with ignore_case flag not being respected

• Fixed a hang in speaker diarization

• Fixed an error related to paths ending in trailing slashes which caused MFA to try to connect to a database named after the local user

• Partial migration to using pathlib.Path instead of os.path

2.1.5

• Fix for improperly reset databases

2.1.4

• Change how database connections are made to remove pooling

2.1.3

• Fixed a bug with intervals after the end of the sound file having negative duration (they are now not parsed)

• Fixed an issue where utterances were not properly assigned to the correct channels

• Modified the logic for connections to attempt to solve error with too many clients

2.1.2

• Fixed a crash in training when the debug flag was not set

• Set default postgresql port to 5433 to avoid conflicts with any system installations

• Fixed a crash in textgrid export

2.1.1

• Fixed a bug with mfa command not working from the command line

• Updated to be compatible with PraatIO 6.0

2.1.0

• Drop support for SQLite as a database backend

• Fixed a bug where TextGrid parsing errors would cause MFA to crash rather than ignore those files

• Updated CLI to use click rather than argparse

• Added --use_phone_model flag for mfa align and mfa validate commands. See Phone model alignments for more details.

• Added --phone_confidence flag for mfa validate commands. See Phone model alignments for more details.

• Added modeling of cutoff phones via --use_cutoff_model which adds progressive truncations of the next word, if it’s not unknown or a non-speech word (silence, laughter, etc). See Modeling cutoffs and hesitations for more details.

• Added support for using SpeechBrain’s VAD model in Segment transcribed files (mfa segment)

• Overhaul and update Train an ivector extractor (mfa train_ivector)

• Overhaul and update Speaker diarization (mfa diarize_speakers)

• Added support for using SpeechBrain’s SpeakerRecognition model in Speaker diarization (mfa diarize_speakers)