Brazilian Portuguese diction for singers: vowels, nasals, t and d, and the r

Sung Brazilian Portuguese has a written standard: a history, vowel rules that differ from speech, and a short list of consonant traps^1,6. This page sits in the diction section. The symbols are explained in IPA for singers, and the neighboring language is covered in Spanish diction.

Where the sung norms come from

The First Congress of the National Language as Sung in 1937 was a first attempt to standardize the diction for Brazilian Portuguese in lyric singing¹. The norms in the proceedings of that congress, published in 1938, grew from a draft by Mário de Andrade⁵. The norms created in 1937 were never officially revised until the 2005 meeting².

The sources disagree on the English date. The University of Maryland guide says the norms were revised and published in both Portuguese and English in 2007². The Indiana document, the ICVT paper and the Revista Música Hodie article date the English version to 2008, a year after the Portuguese^6,3,5. The later, more detailed dating wins: Portuguese 2007, English 2008.

The 2005 vote proposed a phonetic table based on the IPA and a neutral Portuguese pronunciation, avoiding regionalisms and accents³. The goal is a recognizably Brazilian standard for classical singing, without foreignisms or regionalisms⁷. The 2007 norms were meant to be accessible to national and foreign singers alike¹. The 1938 norms are available in an English translation in a 2023 Indiana University doctoral document with an updated critical edition⁶. A University of Maryland guide to singing in Brazilian Portuguese covers diction, IPA and repertoire².

Vowels: open and closed

Brazilian Portuguese has five vowel letters with fifteen phonetic sounds. They can be open or closed, oral or nasal³. The 1938 norms list 19 vowels in the language standard⁶. A phonemic distinction is made between the close-mid vowels /e o/ and the open-mid vowels /ɛ ɔ/⁴.

Stressed /a e o/ remain unchanged throughout Brazil⁴. Unstressed /ɛ ɔ/ are raised to /e o/ in the Southern, Central-West and Southeast regions and remain open in the North and Northeast⁴. In stressed syllables the pronunciation of e and of o varies, and the non-Brazilian singer is told to consult a bilingual dictionary². It is important to always consult the dictionary for word stress as well, since most words receive their tonic stress on the penultimate syllable².

In Brazilian Portuguese the unstressed final /o/ and /e/ are pronounced [ʊ] and [ɪ]². They may be pronounced as written, [o] and [e], in southern Brazil². The unstressed a at the end of a word is always [ɐ]³. In an ICVT analysis, the monosyllabic que is sung [ke] by a recorded soprano when it should be [kɪ]³.

Closed vowels, high notes and low notes

Brazilian speech has more closed vowels than open ones⁶. The 1938 norms recommend the tendency toward closed vowels in classical singing, for a more national timbre and greater ease of emission⁶. If the stressed vowel is closed, it closes or mutes the vowel before it, and if open, it opens it⁶.

Open oral vowels on high to very high notes are closed discreetly as a compromise⁶. Closed vowels on very low notes are opened discreetly, for greater sonority and vocal projection⁶. The unstressed a is reduced, and on sustained notes it is closed discreetly in the middle and upper voice and opened in the lower voice⁶. The final non-nasal e is voiceless, sounding approximately as i⁶. The e of non-final post-tonic syllables is always closed⁶.

Nasal vowels

There are five nasal vowels in Brazilian Portuguese: [ɐ̃] [ẽ] [ĩ] [õ] and [ũ]². There are no open nasal vowels². A vowel nasalizes when a tilde is added above it, which happens only on a and o, or when it is followed by m or n³. Campo is [ˈkɐ̃.pʊ]³. Vowels followed by the digraph nh are always nasal². At the end of a syllable or word, the letter m is related to nasalization of the preceding vowel².

There is no closing of the lips at the end of the articulation of /m/ or /n/ in final position³. When the syllables em and im, the ends of oxytone words, are followed by a word beginning with a vowel, the m becomes the consonantal value nh⁶, the phoneme [ɲ]⁵. In Brazil and Angola the sound /ɲ/ is realized as a nasal palatal approximant [j̃], which nasalizes the vowel that precedes it⁴.

Diphthongs and the semivowel

Singers treat a diphthong as a single vowel followed by a semivowel, or preceded by a semi-consonant⁶. The semivowel is executed with the same discretion and brevity as the consonants⁶. The vowels carry approximately all of the duration of musical sounds⁶.

Nearly all of Brazil pronounces the final stressed i as a falling diphthong in speech⁶. That tendency is not observed in singing, neither on rapid nor slow notes, because it would cause an inelegant and damaging oscillation to the sung voice⁶.

Consonants

Consonants must be articulated with perfect clarity, but always with the greatest possible brevity and discretion⁶. The 1938 norms list 25 consonants in the language standard⁶. In a cluster, the first consonant of the group is conceived as a very quick and clear acciaccatura⁶.

t and d before i

In most Brazilian dialects t and d are pronounced [tʃ] and [dʒ] before [i], [ĩ], consonants and at the end of words⁴. Gente is [ˈʒẽ.tʃɪ]³, and jardim is [ʒar.ˈdʒĩ]². The dental t is articulated as in Latin languages and not aspirated as in English².

s, z, lh, nh, j and h

The initial s is always [s], and is pronounced [z] only in an intervocalic position³. At the end of a syllable, s is [z] if followed by a voiced consonant². The 1938 norms omit the chiado, the j and x sounds, for an s or z before a consonant⁶. A final s or z before a vowel-initial word is [z]⁵, and before b, d, g, l, m, n, r, v or z it sounds clearly as z⁶.

The letter j is always [ʒ]². The lh is always [ʎ], as in bolha [ˈbo.ʎɐ]², and the two letters merge to form an indivisible sound⁶. The nasal palatal [ɲ] is articulated with the body of the tongue against the roof of the mouth². In sloppy speech nh converts to n, as in compania for companhia, and that is avoided in classical singing⁶. The letter h is silent at the beginning of a word, as in hora [ˈɔ.rɐ]². The uncultured swap of v for b, as in berruga for verruga, is rejected in classical singing⁶.

Final l

A final l should never be vocalized, as in animau for animal⁶. The singer should also avoid the contrary vice of rolling the l to the point of duplicating it⁶.

The r

The r is the consonant most open to discussion in Brazilian Portuguese². The rolled /r/ must never be overstressed as it is in Italian², and its articulation should be light³. A lightly rolled r is considered appropriate for all occasions². When in doubt, roll it².

Final r is a very light r, almost nonexistent, never omitted outright⁶. Excessively rolling the final r and giving it a guttural value is rejected⁶. A strong rr in the consonant clusters br, dr, fr, gr, pr and tr (brruto for bruto) is a foreignism the norms repudiate⁶. Once a choice of pronunciation is made, it must be consistent throughout the piece³. The University of Maryland guide points to a simple flap for consonant clusters and the intervocalic position².

Final consonants and paragoge

In words ending in consonants that are not l, m, n, r, s or z, the customary paragoge of the language is used: sôbi for sob⁶. If the note is short, it is divided into two notes of equal duration⁶. Sob before a consonant is not pronounced with a support vowel⁵.

The inserted [ɪ] of epenthesis is a pitfall because it is never written in and is practically imperceptible to the ear³. It appears in clusters, as in submeter [su.bɪ.me.ˈter]³, and before a word-final fricative, as in dez [dɛːɪs]³. The schwa and the nasal vowels are the other pitfalls named in the ICVT paper for non-native singers³.

Word junctions

When a word ends in a stressed vowel, both vowels across the boundary are generally kept⁵. A stressed final vowel meeting a stressed initial one is normally kept as a hiatus⁵. Two unstressed vowels at a word boundary may stay in hiatus, become a diphthong, lengthen, or be elided⁵. Where no vowel is omitted, one vowel is treated as real and the others as semivowels or semi-consonants⁵.

Diphthongization joins the final vowel of one word with the initial vowel of the next when one of the vowels is high⁷. In the eleven score passages that fit that definition, free variation between diphthongization and elision is excluded: the singer must diphthongize⁷. When two notes are written for two vowels, one for each, no word-junction process applies and the vowels stay separate⁷. Post-vocalic /l/ before a vowel-initial word also forms a diphthong in Brazilian Portuguese⁷. In degemination, two identical sounds behave as one⁷. In a corpus of recordings, the nine degeminations were pronounced according to the norm⁷.

Set expressions with the preposition de, such as the apocopes consecrated by national usage, are accepted even when the text does not show them⁶. Assimilation, syncope, metathesis and similar phenomena are normally avoided in classical singing when not required by the text⁶.

Sung versus spoken

In Brazil the denasalization in pão turns it into pau in speech⁷. In classical singing, the foreignisms of musical emission and diction in Brazilian singing are strongly repudiated⁶.

European Portuguese is the contrast case. Its most striking difference from Brazilian Portuguese is vowel reduction², and its /l/ is velarized [ɫ] in all positions⁴. A study of songs by Francisco Mignone addresses a parallel between European and Brazilian Portuguese pronunciation, showing specific differences¹. The University of Nebraska-Lincoln dissertation applies Brazilian diction to selected Mignone songs¹.

To turn these rules into a working score, see transcribe an aria into IPA. For the language whose roll the r must not imitate, see Italian diction².

Common mistakes

• Using a schwa. There is no schwa [ə] in sung Portuguese³.
• Opening final e and o. In Brazilian Portuguese they are [ɪ] and [ʊ]².
• Singing que as [ke]. The monosyllable should be [kɪ]³.
• Skipping the nasal. Five nasal vowels exist, and ão needs the nasal diphthong³.
• Hard t and d before i. They are [tʃ] and [dʒ]⁴.
• Over-rolling the r. The rolled r must never be exaggerated³.
• Leaving the r inconsistent. Keep one choice throughout the piece³.
• Vocalizing a final l. Animau for animal is avoided⁶.
• Turning nh into n. Compania for companhia is avoided⁶.
• Letting a spoken diphthong into the sung line. The falling diphthong on final stressed i stays out of singing⁶.