Welcome to Phonemizer’s documentation!

  • phonemizer allows simple phonemization of words and texts in many languages.

  • Provides both the phonemize command-line tool and the Python function phonemizer.phonemize().

  • It is based on four backends: espeak, espeak-mbrola, festival and segments. The backends have different properties and capabilities resumed in table below. The backend choice is let to the user.

    • espeak-ng is a Text-to-Speech software supporting a lot of languages and IPA (International Phonetic Alphabet) output.

    • espeak-ng-mbrola uses the SAMPA phonetic alphabet instead of IPA but does not preserve word boundaries.

    • festival is another Text-to-Speech engine. Its phonemizer backend currently supports only American English. It uses a custom phoneset, but it allows tokenization at the syllable level.

    • segments is a Unicode tokenizer that build a phonemization from a grapheme to phoneme mapping provided as a file by the user.

espeak

espeak-mbrola

festival

segments

phone set

IPA

SAMPA

custom

user defined

supported languages

100+

35

US English

user defined

processing speed

fast

slow

very slow

fast

phone tokens

yes

yes

yes

yes

syllable tokens

no

no

yes

no

word tokens

yes

no

yes

yes

punctuation preservation

yes

no

yes

yes

stressed phones

yes

no

no

no

tie

yes

no

no

no

To reference phonemizer in your own work, please cite the following JOSS paper.

@article{Bernard2021,
  doi = {10.21105/joss.03958},
  url = {https://doi.org/10.21105/joss.03958},
  year = {2021},
  publisher = {The Open Journal},
  volume = {6},
  number = {68},
  pages = {3958},
  author = {Mathieu Bernard and Hadrien Titeux},
  title = {Phonemizer: Text to Phones Transcription for Multiple Languages in Python},
  journal = {Journal of Open Source Software}
}