graFone pt|en or here.

Grapheme-to-Phoneme (G2P) Converter for European Portuguese

For a description, see here.

vErbuS pt|en

Conjugation and pronunciation system of verbs in European Portuguese

For a description, see here.

TalkIt Demos in Video


The LetsRead Database

- LetsReadDB : 20 hours of reading speech of 284 children.


Pronunciation Dictionaries

- voc_CETEMP_40k.txt : the 41586 most frequent Portuguese words in CETEMPúblico corpus.

Dictionaries in SAMPA with and without: 1) vowel stress marking, 2) symbols for digraphs, 3) alignment with the graphemes:

- dic_CETEMP_50k_mc_v9.txt - The most recent dictionary for continuous speech, with semivowels and vowel stress marking in SAMPA; some alternate pronunciations (separated with ";").
- dic_CETEMP_50k_ipa_v9.txt - Same in IPA.

- dic_CETEMP_50k_mc_v6.txt - A version with few semivowels, in SAMPA.

- dic_CETEMP_40k_acentuado.txt (with vowel stress marking)
- dic_CETEMP_40k_alinhado.txt (with alignment between graphemes and phonemes; uses single-char symbols for phonemes according to a convention).
- dic_CETEMP_40k_alinhado_digrafos.txt (with grapheme-symbols for digraphs).
- dic_CETEMP_40k_alinhado_digrafos_nao_nasais.txt (not using nasal digraphs: an, am, en,...,um).
- dic_CETEMP_40k_acentuado_alinhado_digrafos.txt (with vowel stress marking and symbols for digraphs).

- Minimal Pairs: pairs of words in Portuguese, which differ in only one phoneme. Each line in this file has a pair of words followed by the corresponding pair of transcriptions. A more complete list is here.


Grapheme-to-Phoneme Converter in Portuguese

- Executables (win) in 32 and 64 bits to convert a given voc. file to a pronunciation dic. and to train a model given a dic. file.
- n-gram models (for n=2 to 8) for dic_CETEMP_40k_acentuado_alinhado_digrafos.txt (open folder; the first lines of these files show the possible pronunciations of each grapheme).
- Source code here.


Software (Windows) to record utterances from multiple microphones

- Records utterances from prompted sentences/commands shown on the screen.
- Speakers can read the sentences "Off-air", before starting the utterance recording ("On-air").
- Requires .NET Framework 3.5.
- The .msi installer is here.


Soundproof Booth

Our lab has a soundproof booth (Absorsor ABSLOC.15 cabin with 130x124x250 cm) for high quality recordings of speech. If you need to use this facility, please contact us.

Soundproof Booth


@2016 - Instituto de Telecomunicações