Welcome to Montreal Forced Aligner’s documentation!

Contents:

• Introduction
  • What is forced alignment?
  • Montreal Forced Aligner
    • Pipeline of training
    • Use of speaker information
    • Underlying technology
  • Other forced alignment tools
  • Contributors
  • Citation
  • Funding
• Installation
  • Mac
  • Windows
  • Linux
  • Building from source
  • Files created when using the Montreal Forced Aligner
• Data preparation
  • Dictionaries
    • Non-speech annotations
  • Sound files
    • Sampling rate
    • Duration
  • Data formats
    • Prosodylab-aligner format
    • TextGrid format
    • Transcription normalization and dictionary lookup
• Running the aligner
  • Common options for both aligner executables
  • Align using pretrained models
  • Align using only the data set
• Example
  • Set up
  • Alignment
    • Aligning using pre-trained models
    • Aligning through training
• Common errors (and how to fix them)
  • <unk> words in the output
• Grapheme-to-Phoneme (G2P)
  • Generating a dictionary
    • Use
    • Example
  • Training a new G2P model
    • Use
    • Example
• Pretrained models
  • Pretrained acoustic models
  • Pretrained G2P models
• Changelog
  • 1.0.0
  • 0.8.0
  • 0.7.3
  • 0.7.2
  • 0.7.1
  • 0.7.0
  • 0.6.3
  • 0.6.2
  • 0.6.1
  • 0.6.0
  • 0.5.0
• API Reference
  • Aligner API
    • PretrainedAligner
    • TrainableAligner
  • Corpus API
    • Corpus
  • Dictionary API
    • Dictionary
  • Model API
    • AcousticModel
    • G2PModel
  • Multiprocessing API
    • mfcc
    • compile_train_graphs
    • mono_align_equal
    • align
    • acc_stats
    • tree_stats
    • calc_fmllr
    • convert_alignments
    • convert_ali_to_textgrids
  • Configuration API
    • MfccConfig
    • MonophoneConfig
    • TriphoneConfig
    • TriphoneFmllrConfig

Indices and tables

• Index
• Module Index
• Search Page