Standard Serbo-Croatian has 35 phonemes including vowel length (30 according to traditional analysis which doesn't take vowel length into account): 25 consonants and 10 vowels (5 according to traditional analysis), and a pitch accent, whereas Montenegrin has two more consonants.
/v/ is a phonetic fricative, although it has less frication than /f/. However, it does not interact with unvoiced consonants in clusters as a fricative would, and so is considered to be phonologically a sonorant (approximant).
/t͡s, f, x/ are voiced [d͡z, v, ɣ] before voiced consonants.
Glottal stop [ʔ] may be inserted between vowels across word boundary, as in i onda[iː ʔônda].
/ʂ, ʐ/ are retracted to [ɕ, ʑ] before /t͡ɕ, d͡ʑ/.
/x/ is retracted to [h] when it is initial in a consonant cluster, as in hmelj[hmêʎ].
/ʋ/ is labiovelar [w] before /u/, as in vuk[wûːk].
/r/ can be syllabic, short or long, and carry rising or falling tone, e.g. kȓv ('blood'), sȑce ('heart'), sŕna ('deer'), mȉlosr̄đe ('compassion'). It is typically realized by inserting a preceding or (more rarely) succeeding non-phonemic vocalic glide.
/l/ is generally velarized or "dark" [ɫ]. Diachronically, it was fully vocalized into /o/ in coda positions, as in past participle *radil > radio ('worked'). In some dialects, notably Torlakian, that process did not take place, and /l/ can be syllabic as well. However, in the standard language, vocalic /l/ appears only in loanwords, as in the name for the Czech river Vltava for instance, or debakl, bicikl. Very rarely other sonorants are syllabic, such as /ʎ̩/ in the surname Štarklj and /n̩/ in njutn ('newton').
The retroflex consonants /ʂ, ʐ, tʂ, dʐ/ are, in more detailed phonetic studies, described as apical[ʃ̺, ʒ̺, t̺ʃ̺ʷ, d̺ʒ̺ʷ]. In most spoken Croatian idioms, as well as in some Bosnian, they are postalveolar (/ʃ, ʒ, t͡ʃ, d͡ʒ/) instead, and there could be a complete or partial merger between /tʂ, dʐ/ and palatal affricates /tɕ, dʑ/.
Voicing contrasts are neutralized in consonant clusters, so that all obstruents are either voiced or voiceless depending on the voicing of the final consonant, though this process of voicing assimilation may be blocked by syllable boundaries.
Vowel space of Serbo-Croatian from Landau et al. (1999:67). The diphthong /ie/ occurs in some Croatian and Serbian dialects. Schwa[ə] only occurs allophonically.
The Serbo-Croatian vowel system is symmetrically composed of five vowel qualities /a, e, i, o, u/.  Although the difference between long and short vowels is phonemic, it is not represented in standard orthography, as it is in Czech or Slovak orthography, except in dictionaries. Unstressed vowels are shorter than the stressed ones by 30% (in the case of short vowels) and 50% (in the case of long vowels).
The long Ijekavian reflex of Proto-Slavicjat is of disputed status. The prescriptive grammar Barić et al. (1997) published by the foremost Croatian normative body—the Institute of Croatian Language and Linguistics, describes it as a diphthong, but this norm has been heavily criticized by phoneticians as having no foundation in the spoken language, the alleged diphthong being called a "phantom phoneme". Thus the reflex of long jat, which is spelled as a trigraph ⟨ije⟩ in standard Croatian, Bosnian and Ijekavian Serbian, represents the sequence /jeː/.
Stressed vowels carry one of the two basic tones, rising and falling.
New Shtokavian dialects (which form the basis of standard language) allow two tones on stressed syllables and have distinctive vowel length and so distinguish four combinations, called pitch accent: short falling (ȅ), short rising (è), long falling (ȇ), and long rising (é).
Most speakers from Serbia and Croatia do not distinguish between short rising and short falling tones. They also pronounce most unstressed long vowels as short, with some exceptions, such as genitive plural endings.
The accent is relatively free because it can be on any syllable except the last one. (Note, however, that the word paradàjz ('tomato' nominative sg.) normally bears a short rising tone on the final syllable in the speech of educated speakers. Further exceptions to this prescriptive rule are found in words such as fabrikànt ('manufacturer' nominative sg.), asistènt ('assistant' nominative sg.), apsolvènt ('student who has fulfilled all requirements except an honours thesis' nominative sg.), trafikànt ('sales assistant at a newsstand' nominative sg.), etc.) This is relevant for Serbia, where educated speakers otherwise speak close to standard Serbian; this is less so in Croatia, where educated speakers often use a local Croatian variant which might have a quite different stress system. For example, even highly educated speakers in Zagreb will have no tones, and can have stress on any syllable.
Accent alternations are very frequent in inflectional paradigms, in both quality and placement in the word (the so-called "mobile paradigms", which were present in Proto-Indo-European itself and became much more widespread in Proto-Balto-Slavic). Different inflected forms of the same lexeme can exhibit all four accents: lònac/ˈlǒnats/ ('pot' nominative sg.), lónca/ˈlǒːntsa/ (genitive singular), lȏnci/ˈlôːntsi/ (nominative plural), lȍnācā/ˈlônaːtsaː/ (genitive plural).
Research done by Pavle Ivić and Ilse Lehiste has shown that all stressed syllables of Serbo-Croatian words are basically spoken with a high tone and that native speakers rely on the phonetic tone of the first post-tonic syllable to judge the pitch accent of any given word. If the high tone of the stressed syllable is carried over to the first post-tonic syllable, the accent is perceived as rising. If it is not, the accent is perceived as falling, which is the reason monosyllabic words are always perceived as falling. Therefore, truly narrow phonetic transcriptions of lònac, lónca, lȏnci and lȍnācā are [ˈlónáts, ˈlóːntsá, ˈlóːntsì, ˈlónàˑtsàˑ] or the equivalent [ˈlo˦nats˦, ˈloːn˦tsa˦, ˈloːn˦tsi˨, ˈlo˦naˑ˨tsaˑ˨]. Ivić and Lehiste were not the first scholars to notice this; in fact, Leonhard Masing made a very similar discovery decades earlier, but it was ignored due to him being a foreigner and because it contradicted the Vukovian approach, which was then already well-ingrained.
Although distinctions of pitch occur only in stressed syllables, unstressed vowels maintain a length distinction. Pretonic syllables are always short, but posttonic syllables may be either short or long. These are traditionally counted as two additional accents. In the standard language, the six accents are realized as follows:
Examples are short falling as in nȅbo ('sky') /ˈnêbo/; long falling as in pîvo ('beer') /ˈpîːvo/; short rising as in màskara ('eye makeup') /ˈmǎskara/; long rising as in čokoláda ('chocolate') /t͡ʂokoˈlǎːda/. Unstressed long syllables can occur only after the accented syllable, as in d(j)èvōjka ('girl') /ˈd(ј)ěvoːjka/ or dòstavljānje ('delivering') /ˈdǒstavʎaːɲe/. There can be more than one post-accent length in a word, notably in genitive plural of nouns: kȍcka ('cubes') → kȍcākā ('cubes''). Realization of the accents varies by region.
Restrictions on the distribution of the accent depend, beside the position of the syllable, also on its quality, as not every kind of accent can be manifested in every syllable.
Falling tone generally occurs in monosyllabic words or the first syllable of a word (pȃs ('belt'), rȏg ('horn'); bȁba ('old woman'), lȃđa ('river ship'); kȕćica ('small house'), Kȃrlovac). The only exception to this rule are interjections, words uttered in the state of excitement (such as ahȁ, ohȏ)
Thus, monosyllabics generally have falling tone, and polysyllabics generally have falling or rising tone on the first syllable and rising in all the other syllables but the last one. The tonal opposition rising ~ falling is hence generally possible only in the first accented syllable of polysyllabic words, and the opposition by lengths, long ~ short, is possible even in the non-accented syllable as well as in the postaccented syllable (but not in the preaccented position).
Proclitics, clitics that latch on to a following word, on the other hand, may "steal" a falling tone (but not a rising tone) from the following monosyllabic or disyllabic word. The stolen accent is always short and may end up being either falling or rising on the proclitic. The phenomenon (accent shift to proclitic) is most frequent in the spoken idioms of Bosnia, as in Serbian it is more limited (normally with the negation proclitic ne) and it is almost absent from Croatian Neo-Shtokavian idioms. Such a shift is less frequent for short rising accents than for the falling one (as seen in this example: /ʒěliːm/ → /ne ʒěliːm/).
'I don't want'
'in the winter'
'not being able to'
'I can't see'
'to the city' (stays falling)
'in the forest' (becomes rising)
Serbo-Croatian exhibits a number of morphophonological alternations. Some of them are inherited from Proto-Slavic and are shared with other Slavic languages, and some of them are exclusive to Serbo-Croatian, representing later innovation.
The so-called "fleeting a" (Serbo-Croatian: nepóstojānō a), or "movable a", refers to the phenomenon of short /a/ making apparently random appearance and loss in certain inflected forms of nouns. This is a result of different types of reflexes Common Slavic jers */ъ/ and */ь/, which in Štokavian and Čakavian dialects merged to one schwa-like sound, which was lost in a weak position and vocalized to */a/ in a strong position, giving rise to what is apparently unpredictable alternation. In most of the cases, this has led to such /a/ appearing in word forms ending in consonant clusters, but not in forms with vowel ending.
The "fleeting a" is most common in the following cases:
in nominative singular, accusative singular for inanimate nouns, and genitive plural for certain type of masculine nouns:
The output of the second and the third Slavic palatalization is in the Serbo-Croatian grammar tradition known as "sibilantization" (sibilarizácija/сибилариза́ција). It results in the following alternations before /i/:
/k/ → /ts/
/ɡ/ → /z/
/x/ → /s/
This alternation is prominently featured in several characteristic cases:
in the imperative forms of verbs with stem ending in /k/, /ɡ/ and one verb in /x/:
There are two types of consonant assimilation: by voicing (jednačenje po zvučnosti) and by place of articulation (jednačenje po m(j)estu tvorbe).
Assimilation of voice
All consonants in clusters are neutralized by voicing, but Serbo-Croatian does not exhibit final-obstruent devoicing as most other Slavic languages do. Assimilation is practically always regressive, i.e. voicing of the group is determined by voicing of the last consonant. Sonorants are exempted from assimilation, so it affects only the following consonants:
/b/ ↔ /p/
kobac ('hawk') →kobca : kopca (nominative → genitive, with fleeting a)
top ('cannon') + džija → topdžija : tobdžija ('cannonman')
Furthermore, /f/, /x/ and /ts/ don't have voiced counterparts, so they trigger the assimilation, but are not affected by it.
As can be seen from the examples above, assimilation is generally reflected in orthography. However, there are numerous orthographic exceptions, i.e. even if voicing or devoicing does take place in speech, the orthography does not record it, usually to maintain the etymology clearer.
Assimilation by place of articulation
Assimilation by place of articulation affects /s/ and /z/ in front of (post)alveolars /ʃ/, /ʒ/, /t͡ʂ/, /d͡ʐ/, /tɕ/, /dʑ/, as well as palatals /ʎ/ and /ɲ/, producing /ʃ/ or /ʒ/:
A historical /l/ in coda position has become /o/ and is now so spelled, and produces an additional syllable. For example, the Serbo-Croatian name of Belgrade is Beograd. However, in Croatian, the process is partially reversed; compare Croatian stol, vol, sol vs. Serbian sto, vo, so ('table', 'ox' and 'salt').
The sample text is a reading of the first sentence of The North Wind and the Sun by a 57-year-old female announcer at the Croatian Television Network reading in a colloquial style.
^Stevanović, Mihailo (1986). Савремени српскохрватски језик. Belgrade: Naučna knjiga. p. 82. И при изговору сугласника ж и ш [...] врх се језика диже према предњем делу предњег непца, и овлаш га додирује на делу одмах иза алвеола.
^Barić et al. (1997:49) "Prednji je i složeni samoglasnik, dvoglasnik (diftong) ie. Pri njegovu su izgovoru govorni organi najprije u položaju sličnom kao pri izgovoru glasa i, a onda postupno prelaze u položaj za izgovor glasa e. U hrvatskom književnom jeziku dvoglasnik je ie ravan diftong."
^Kapović (2007:66) "Iako se odraz dugoga jata u kojem ijekavskom govoru možda i može opisati kao dvoglas, on tu u standardu sasma sigurno nije. Taj tobožnji dvoglas treba maknuti iz priručnikâ standardnoga jezika jer nema nikakve koristi od uvođenja fantomskih fonema bez ikakve podloge u standardnojezičnoj stvarnosti."
^This is a stylistically marked form: the usual plural form of vrȃg is with -ov- interfix: vrȁgovi; accusative plural: vrȁgove, but the infix is inhibiting the environment conditioning the palatalization, so the short plural form was provided.
Gick, Bryan; Campbell, Fiona; Oh, Sunyoung; Tamburri-Watt, Linda (2006), "Toward universals in the gestural organization of syllables: A cross-linguistic study of liquids", Journal of Phonetics, Vancouver: Department of Linguistics, University of British Columbia, 34 (1): 49–72, doi:10.1016/j.wocn.2005.03.005
Jazić, Đorđe (1977), Osnovi fonetike ruskog jezika: ruski glasovni sistem u poredjenju sa srpskohrvatskim, Beograd: Naučna knjiga
Jovanović Maldoran, Srđan (2014). "Prilog izučavanja akcenatskog kvaliteta i kvantiteta srpske varijante policentričnog srpskohrvatskog jezika" [To the study of Accentual Quality and Quantity of Serbian Version of the Polycentric Serbo-Croatian Language]. Slavia : časopis pro slovanskou filologii (in Serbo-Croatian). Prague. 83 (2): 179–185. ISSN0037-6736. ZDB-ID204528-x.
Landau, Ernestina; Lončarić, Mijo; Horga, Damir; Škarić, Ivo (1999), "Croatian", Handbook of the International Phonetic Association: A guide to the use of the International Phonetic Alphabet, Cambridge: Cambridge University Press, pp. 66–69, ISBN0-521-65236-7
Lehiste, Ilse; Ivić, Pavle (1963), Accent in Serbocroatian : an experimental study, University of Michigan, Dept. of Slavic Languages and Literatures
Lehiste, Ilse; Ivić, Pavle (1986), Word and sentence prosody in Serbocroatian, MIT Press, ISBN0262121115