Library for manipulating International Phonetic Alphabet (IPA) pronunciations.

Features include:

• Getting the category and details of a phone, e.g. "open front unrounded vowel" for ɶ
• Splitting IPA pronunciations into groups of:
  • Phones (/ˈt͡ʃuːz/ to ˈt͡ʃ uː z )
  • Phonemes (/kˈaʊ/ to k ˈaʊ for U.S. English)
• Converting pronunciations between:
  • IPA
  • espeak
  • sampa

Supported Languages:

• Arabic (ar)
• Czech (cs-cz)
• German (de-de)
• U.S. English (en-us)
• U.K. English (en-gb)
• Spanish (es-es)
• Persian/Farsi (fa)
• Spanish (es-es)
• Italian (it-it)
• Luxembourgish (lb-lb)
• Dutch (nl)
• Portuguese (pt)
• Russian (ru-ru)
• Swahili (sw)

$ pip install gruut-ipa

• Python 3.6 or higher

For command-line usage, you may also want:

• espeak
• jq

Install these with:

$ sudo apt-get install espeak jq

Phones in IPA are composed of different components:

• Letters
  • Non-combining Unicode characters that represent a distinct human sound (phone)
• Suprasegmentals
  • Non-combining Unicode characters that represent language features above individual vowels or consonants
  • Stress (ˈˌ), elongation (ː), linking/ties (t͡s), and short/long breaks (| ‖) are suprasegmentals
• Diacritics
  • Combining characters that provide additional information about a phone's pronunciation, such as nasalation

See IPA Chart for more details.

While phones represent individual sounds, phonemes are the phonetic units of a language that meaningfully distinguish words. A phoneme may be realized by many different phones. For example, the /r/ in Standard German can be realized as a uvular fricative (χ/ʁ), a uvular approximant (ɹ), or a uvular tap or trill (ʀ/r).

A phoneme may also be composed of multiple phones, such as the dipthong aʊ in U.S. English (the "ow" in "cow").

Supported languages in gruut-ipa contain a phonemes.txt file in the gruut_ipa/data directory. This file has the following format:

<phoneme> <example> [<replace> ...]

where <phoneme> is a set of IPA letters, like ɶ or aʊ. The <example> is a word whose pronunciation contains the <phoneme>. After that, there are one or more optional <replace> strings that will be replaced with <phoneme>. The German /r/ example from above might be represented as:

r brot χ ʁ ɹ ʀ

Phonemes for a given language come from phonological analyses and from public databases. Ultimately, they are geared towards capturing pronunciations from Wiktionary.

Print JSON information about phones:

$ python3 -m gruut_ipa describe "ˈãː" | jq .
{
  "text": "ˈãː",
  "letters": "a",
  "stress": "primary",
  "height": "open",
  "placement": "front",
  "rounded": false,
  "type": "Vowel",
  "nasalated": true,
  "elongated": true
}

Split an IPA pronunciation into phones:

$ python3 -m gruut_ipa phones "ˈjɛs|ˈt͡ʃuːz aɪpiːeɪ‖"
ˈj ɛ s | ˈt͡ʃ uː z a ɪ p iː e ɪ ‖

Group phones into phonemes for a specific language:

$ python3 -m gruut_ipa phonemes en-us "/dʒʌst ə kaʊ/"
d͡ʒ ʌ s t ə k aʊ

Convert between IPA, espeak, and sampa:

$ python3 -m gruut_ipa convert ipa espeak "mʊmˈbaɪ"
[[mUm'baI]]

$ python3 -m gruut_ipa convert espeak ipa "[[D,Is Iz sVm f@n'EtIk t'Ekst 'InpUt]]"
ðˌɪs ɪz sʌm fɘnˈɛtɪk tˈɛkst ˈɪnpʊt

Chain commands together:

$ python3 -m gruut_ipa convert espeak ipa "[[k'aU]]" | \
    python3 -m gruut_ipa phonemes en-us --keep-stress
k ˈaʊ

Some languages have multiple phoneme sets available:

• U.S. English (en-us)
  • CMUDict (en-us/cmudict)
  • Zamia (en-us/zamia)
• Swahili (sw)
  • ALFFA (sw/alffa)

Convert from IPA to alternative phoneme set:

$ python3 -m gruut_ipa convert ipa en-us/cmudict "h ɛ l ˈoʊ w ˈɚ l d"
HH EH0 L OW1 W ER1 L D

Convert from alternative phoneme set to IPA:

$ python3 -m gruut_ipa convert en-us/cmudict ipa "HH EH0 L OW1 W ER1 L D"
h ɛ l ˈoʊ w ˈɚ l d

Use the speak-ipa script to have espeak pronounce IPA. You may need to apt-get install espeak first.

$ echo '/hɛloʊ wɝld/' | bin/speak-ipa en-us -s 60 -w 'hello world.wav'
$ aplay 'hello world.wav'

Supported IPA phones can be printed with:

$ python3 -m gruut_ipa print
{"text": "i", "letters": "i", "stress": "none", "height": "close", "placement": "front", "rounded": false, "type": "Vowel", "nasalated": false, "elongated": false, "description": "close front unrounded vowel", "espeak": "i", "sampa": "i"}
{"text": "y", "letters": "y", "stress": "none", "height": "close", "placement": "front", "rounded": true, "type": "Vowel", "nasalated": false, "elongated": false, "description": "close front rounded vowel", "espeak": "y", "sampa": "y"}
...

A nice table can be generated with jq:

$ python3 -m gruut_ipa print | \
    jq -r '. | "\(.text)\t\(.espeak)\t\(.sampa)\t\(.description)"'

Converted to Markdown:

If you see anything wrong or missing, please let me know.