This is a modified implementation of the Penn Phonetic Forced Aligner (P2FA) [1]. Relative to the original implementation, this repo provides the following.

• Support for Python 3
• Support for performing forced alignment both in Python and on the command-line
• Fewer alignment failures due to, e.g., out-of-vocabulary (OOV) words or punctuation
• Direct integration with pypar, a feature-rich phoneme alignment representation.
• Multiprocessing for quickly aligning speech datasets
• Clean, documented code

pyfoal depends on HTK and has been tested on Mac OS and Linux using HTK version 3.4.0. There are known issues in using version 3.4.1 on Linux. HTK is released under a license that prohibits redistribution, so you must install HTK yourself and verify that the commands HCopy and HVite are available as system-wide binaries. After downloading HTK, I use the following for installation on Linux.

sudo apt-get install -y gcc-multilib libx11-dev
sudo chmod +x configure
./configure --disable-hslab
make all
sudo make install

For more help with HTK installation, see notes by Jaekoo Kang and Steve Rubin.

pip install pyfoal

alignment = pyfoal.align(text, audio, sample_rate)

text is a string containing the speech transcript. audio is a 1D numpy array containing the speech audio.

# Return the resulting alignment
alignment = pyfoal.from_file(text_file, audio_file)

# Save alignment to json
pyfoal.from_file_to_file(text_file, audio_file, output_file)

If you need to align many files, use from_files_to_files, which accepts lists of files and uses multiprocessing.

usage: python -m pyfoal
    [-h]
    --text TEXT [TEXT ...]
    --audio AUDIO [AUDIO ...]
    --output OUTPUT [OUTPUT ...]

optional arguments:
  -h, --help            show this help message and exit
  --text TEXT [TEXT ...]
                        The speech transcript files
  --audio AUDIO [AUDIO ...]
                        The speech audio files
  --output OUTPUT [OUTPUT ...]
                        The json files to save the alignments

Tests can be run as follows.

pip install pytest
pytest

[1] J. Yuan and M. Liberman, “Speaker identification on the scotus corpus,” Journal of the Acoustical Society of America, vol. 123, p. 3878, 2008.