Downloads · 30 days
267
24% of all-time downloads
kokoro-ipa-lexicons
beshkenadze/kokoro-ipa-lexicons
kokoro-ipa-lexicons is a machine learning model from beshkenadze. Use it for the machine learning task on the model card, and read the license before you ship it in a product.
IPA pronunciation lexicon with stress marks, formatted for use with Kokoro TTS and other espeak-ng-trained speech synthesis models.
All-time downloads
1.1K
Public
Repo size
58.7 MB
Likes
1
Public
Repo makeup
Click a slice to open those files.
.tsv79.1 MB · 100%
At a glance
- Access
- Public
- Created
- Mar 22, 2026
- Updated
- Mar 22, 2026
- SHA
- a3d069ca
Model card
From the Hugging Face model README
IPA Lexicon — {language}
IPA pronunciation lexicon with stress marks, formatted for use with Kokoro TTS and other espeak-ng-trained speech synthesis models.
Format
Tab-separated values (TSV): word\tphonemes
Phonemes are space-separated IPA symbols with primary (ˈ) and secondary (ˌ) stress marks.
hola ˈo l a
prueba p ɾ u ˈe β a
español ˌe s p a ɲ ˈo l
Usage
For Kokoro TTS, join phonemes by removing spaces. Move stress marks before syllable onsets for best results.
Source & Attribution
This lexicon was extracted from gruut by Michael Hansen (rhasspy), licensed under the MIT License.
Specifically, data comes from the espeak/lexicon.db files in the gruut-lang-* packages (v2.0.x). These contain IPA transcriptions with stress marks generated via espeak-ng.
Licensing Note
- gruut (framework): MIT License
- gruut-lang-* (data packages): Published under the gruut project (MIT), but the
espeak/subdirectory contains phonemizations generated by espeak-ng (GPL v3). The legal status of bulk phonemization output from a GPL tool is debated — it may or may not constitute a derivative work depending on jurisdiction. - This redistribution: Provided as-is under the same terms as the gruut project. Users should evaluate GPL compliance for their own use case.
Languages Available
| Language | Code | Entries | File |
|---|---|---|---|
| Arabic | ar | 82,702 | ar_lexicon.tsv |
| Czech | cs | 43,991 | cs_lexicon.tsv |
| German | de | 279,704 | de_lexicon.tsv |
| English | en | 124,447 | en_lexicon.tsv |
| Spanish | es | 596,995 | es_lexicon.tsv |
| Farsi | fa | 47,143 | fa_lexicon.tsv |
| French | fr | 90,717 | fr_lexicon.tsv |
| Italian | it | 29,388 | it_lexicon.tsv |
| Dutch | nl | 40,084 | nl_lexicon.tsv |
| Portuguese | pt | 65,262 | pt_lexicon.tsv |
| Russian | ru | 533,929 | ru_lexicon.tsv |
| Swedish | sv | 27,192 | sv_lexicon.tsv |
| Swahili | sw | 69,676 | sw_lexicon.tsv |
Conversion
Extracted with:
pip download gruut-lang-{code} --no-deps
tar xzf gruut_lang_{code}-*.tar.gz
sqlite3 espeak/lexicon.db -separator $'\t' \
"SELECT word, phonemes FROM word_phonemes ORDER BY word" > {code}_lexicon.tsv
- Created
- Mar 22, 2026
- Updated
- Mar 22, 2026