Model Card for japanese_mfa
This MFA model is for aligning Japanese speech.
Model Details
Model Description
- Developed by: Michael McAuliffe
- Funded by: N/A
- Model type: Montreal Forced Aligner model
- Language(s) (NLP): Japanese
- License: cc-by-4.0
Uses
Direct Use
This model is intended to be used for forced alignment of Japanese speech. Please see https://montreal-forced-aligner.readthedocs.io/en/latest/user_guide/troubleshooting.html for details on common fixes.
Out-of-Scope Use
This model cannot provide accurate assessments of goodness of pronunciations or provide transcripts as it is trained to be accepting of variation in pronunciation to provide a reasonable alignment for Japanese speech.
Bias, Risks, and Limitations
This model will perform best on the variety of speech that it was trained on. The speakers in the training data are all adult speakers, so child speech alignment may not be accurate.
Recommendations
When using this model on a variety that it was not trained on, better results can be attained by adapting the model to the data to be aligned first. See https://montreal-forced-aligner.readthedocs.io/en/latest/user_guide/workflows/adapt_acoustic_model.html and https://github.com/mmcauliffe/mfa-adaptation for example usage and scripts.
How to Get Started with the Model
Use the code below to get started with the model.
To get started, follow the instructions for installing MFA. To align files using this model, use the mfa align command.
Dictionary Details
Details for japanese_mfa dictionary and G2P model
- Source: wikipron
- Orthography: Kanji and kana
- Phone set: MFA
- Words: 2,239,777
- Phones: 78
- Graphemes: 5,752
IPA chart
Consonants
| Manner | Labial | Labiodental | Alveolar | Palatal | Velar | Uvular | Glottal |
|---|---|---|---|---|---|---|---|
| Nasal | m mː | n nː | ɲ ɲː | ŋ | ɴ ɴː | ||
| Palatalized | mʲ mʲː | ||||||
| Stop | p pː b bː | t tː d dː | c cː ɟ ɟː | k kː ɡ ɡː | ʔ | ||
| Palatalized | pʲ pʲː bʲ | tʲ tʲː dʲ dʲː | |||||
| Affricate | ts tsː dz dzː | tɕ tɕː dʑ dʑː | |||||
| Sibilant | s sː z | ɕ ɕː ʑ | |||||
| Fricative | ɸ ɸː | v | ç | h hː | |||
| Palatalized | ɸʲ ɸʲː | vʲ | |||||
| Approximant | w | j | ɰ̃ | ||||
| Tap | ɾ | ||||||
| Palatalized | ɾʲ |
Vowels
| Front | Near-Front | Central | Near-Back | Back | |
|---|---|---|---|---|---|
| Close | i iː i̥ | ɨ ɨː ɨ̥ | ɯ ɯː ɯ̥ | ||
| Close-Mid | e eː | o oː | |||
| Open-Mid | |||||
| Open | a aː |
Training Details
Training Data
Common Voice Japanese
- Source: https://voice.mozilla.org/en/datasets
- License: CC-0
- Dialects: N/A
- Number of hours: 71.54
- Number of utterances: 56,631
- Number of speakers: 1,364
- Female speakers: 199
- Male speakers: 459
- Unknown speakers: 706
GlobalPhone Japanese
- Source: https://catalogue.elra.info/en-us/repository/browse/ELRA-S0199/
- License: ELRA
- Dialects: N/A
- Number of hours: 33.88
- Number of utterances: 13,103
- Number of speakers: 144
- Female speakers: 44
- Male speakers: 104
- Unknown speakers: 0
Japanese Versatile Speech
- Source: https://sites.google.com/site/shinnosuketakamichi/research-topics/jvs_corpus
- License: CC BY-SA 4.0
- Dialects: N/A
- Number of hours: 30.25
- Number of utterances: 14,994
- Number of speakers: 100
- Female speakers: 51
- Male speakers: 49
- Unknown speakers: 100
LaboroTV Japanese
- Source: https://laboro.ai/activity/column/engineer/eg-laboro-tv-corpus-jp/
- License: LaboroTV Non-commercial
- Dialects: N/A
- Number of hours: 1,915.43
- Number of utterances: 1,536,227
- Number of speakers: 305,246
- Female speakers: 0
- Male speakers: 0
- Unknown speakers: 305,246
MagicData ASR-SJapDuSC
- Source: https://magichub.com/datasets/japanese-scripted-speech-corpus-daily-use-sentence/
- License: CC-BY-NC-ND 4.0
- Dialects: N/A
- Number of hours: 18.71
- Number of utterances: 17,302
- Number of speakers: 37
- Female speakers: 19
- Male speakers: 19
- Unknown speakers: 0
Microsoft Speech Language Translation Japanese
- Source: https://msropendata.com/datasets/54813518-4ea6-4c39-9bb2-b0d1e5f0c187
- License: Microsoft Research Data License
- Dialects: N/A
- Number of hours: 9.85
- Number of utterances: 7,338
- Number of speakers: 112
- Female speakers: 0
- Male speakers: 0
- Unknown speakers: 112
TEDxJP-10K
- Source: https://github.com/laboroai/TEDxJP-10K
- License: Apache 2.0
- Dialects: N/A
- Number of hours: 8.85
- Number of utterances: 9,962
- Number of speakers: 271
- Female speakers: 0
- Male speakers: 0
- Unknown speakers: 271
Training Procedure
Preprocessing
Preprocessing include fixes and orthographic standardization to various corpora.
Training Hyperparameters
- Training regime: Training configuration
Evaluation
Testing Data, Factors & Metrics
Testing Data
N/A
Factors
N/A
Metrics
N/A
Results
N/A
Summary
Technical Specifications
Model Architecture and Objective
HMM-GMM model
Software
This model was trained via the Montreal Forced Aligner.
Citation
BibTeX:
@techreport{mfa_japanese_mfa_acoustic_2026,
author={McAuliffe, Michael and Sonderegger, Morgan},
title={Japanese MFA acoustic model v3.3.0},
address={\url{https://huggingface.co/MontrealCorpusTools/japanese_mfa}},
year={2026},
month={Jun},
}
APA:
McAuliffe, M. & Sonderegger, M. (2026). Japanese MFA acoustic model v3.3.0. Available at https://huggingface.co/MontrealCorpusTools/japanese_mfa.
Contact
For questions and issues, please file an issue either for this model at https://huggingface.co/MontrealCorpusTools/japanese_mfa/discussions or for larger MFA issues at https://github.com/MontrealCorpusTools/Montreal-Forced-Aligner/issues.