The phonology of Welsh is characterised by a number of sounds that do not occur in English and are rare in European languages, such as the voiceless alveolar lateral fricative [ɬ] and several voiceless sonorants (nasals and liquids), some of which result from consonant mutation. Stress usually falls on the penultimate syllable in polysyllabic words, while the word-final unstressed syllable receives a higher pitch than the stressed syllable.

Welsh has the following consonant phonemes:^{[1][2][3][4][5][6]}

Symbols in parentheses are either allophones, or found only in loanwords. The sound /z/ generally occurs in loanwords, e.g. sŵ /zuː/ ('zoo'), although this is usually realised as /s/ in northern accents, e.g. /suː/. The postalveolar affricates /tʃ/ and /dʒ/ occur mainly in loanwords, e.g. tsips /tʃɪps/ ('chips') and jeli /ˈdʒɛli/ ('jelly'), but also in some dialects as developments from /tj/ and /dj/, e.g. /dʒaul/ from diafol /ˈdjavɔl/ ('devil'). The voiceless nasals /m̥ n̥ ŋ̊/ occur mostly word-initially, as a consequence of nasal mutation. These nasals have recently been interpreted as sequences of /m n ŋ/ + /h/.^[7][8] Initial /χw/ is colloquially realised as [ʍ] in the south, e.g. chwech /χweːχ/ ('six') pronounced [ʍeːχ].

[ç] results from /j/ when preceded by /h/, often as a result of h-prothesis of the radical word, e.g. iaith /jai̯θ/ 'language' becomes ei hiaith [ɛi çai̯θ] 'her language'.^[9]

The stops /p t k/ are distinguished from /b d ɡ/ by means of aspiration more consistently than by voicing, as /b d ɡ/ are actually devoiced in most contexts. This devoiced nature is recognised in the spelling of /sp sk/ as ⟨sb sg⟩, although /st/ is orthographically ⟨st⟩ for historical reasons.

The fricatives /v ð/ tend not to be pronounced in certain contexts, e.g. nesaf /nɛsav/ ('next') realised as /ˈnɛsa/ or i fyny /iː ˈvənɨ/ ('up') from mynydd /mənɨð, mənɪð/ ('mountain'). Historically, this occurred so often with the voiced velar fricative that it disappeared entirely from the language. The occurrence and distribution of the phoneme /ʃ/ varies from area to area. Very few native words are pronounced with /ʃ/ by all speakers, e.g. siarad /ˈʃarad/ ('talk'), although it appears in borrowings, e.g. siop /ʃɔp/ ('shop'). In northern accents, it can occur when /s/ precedes /iː j/, e.g. es i /ˈeːʃ i/ ('I went'). In some southern dialects it is produced when /s/ follows /ɪ/ or /iː/, e.g. mis /miːʃ/ ('month'). The voiceless fricative /χ/ is realised as uvular except by some southwestern speakers, who produce the sound in the velar region as [x].

The /r/ phoneme is reportedly pronounced as the voiced uvular fricative [ʁ] by some speakers in Dyfed and Gwynedd, in a pronunciation known as tafod tew ('thick tongue').^[10]

In some dialects of north-western Welsh, the /l/ phoneme is consistently velarised or "dark" ([ɫ], not to be confused with [ɬ]) in all positions,^{[citation needed]} but remains unvelarised or "clear" ([l]) in the south, except in rare exceptions where [ɫ] is found after /d/, e.g. dlos [dɫos] 'pretty'.

The vowel phonemes of Welsh are as follows:^{[1][2][3][4][5][6]}

The vowels /ɨ̞/ and /ɨː/ merged with /ɪ/ and /iː/ in southern dialects, but are retained in northern dialects. In all dialects, the contrast between long and short vowels is found in stressed penultimate syllables of polysyllabic words or in monosyllabic words. Word-final vowels show a contrast between vowel quality rather than length proper, e.g. ysbyty /əsˈbə.tiː/ is realised as [əsˈpə.ti] with final short [i] rather than with final long [iː].

The vowel /ə/ does not occur in the final syllable of words (except a few monosyllabic proclitics). It is always pronounced short except when emphasised in the name of the letter y.^[11]

The long counterpart to short /a/ is sometimes misleadingly transcribed /ɑ/. This is often found in solely quality-distinctive transcriptions to avoid using a length mark. The actual pronunciation of long /a/ is [aː], which makes the vowel pair unique in that there is no significant quality difference. Regional realisations of /aː/ may be [æː] or [ɛː] in north-central and (decreasingly) south-eastern Wales or sporadically as [ɑː] in some southern areas undoubtedly under the influence of English.^[11]

The diphthongs containing /ɨ/ occur only in northern dialects; in southern dialects /ʊɨ/ is replaced by /ʊi/ and /ɨu, əɨ~ɛɨ, ɔɨ, a(ː)ɨ/ are merged with /ɪu, əi~ɛi, ɔi, ai/. There is a general tendency in the South to simplify diphthongs in everyday speech, e.g. Northern /ɡwaːɨθ/ corresponding to /ɡwaːθ/ in the South, or Northern /ɡwɛiθjɔ/ and Southern /ɡwiθɔ/.

The long vowels are not inherited from Proto-Celtic vowels, which were transformed into separate vowels, but instead from the New Quantity System.^[12]

Stress in polysyllabic words occurs most commonly on the penultimate syllable, more rarely on the final syllable (e.g. verbs ending in -áu).^[13] Exceptions can arise in relation to borrowings from foreign words, such as ambiwlans and testament (both stressed on the first syllable). According to its positioning, related words or concepts (or even plurals) can sound quite different, as syllables are added to the end of a word and the stress moves correspondingly:

Note also how adding a syllable to ysgrifennydd to form ysgrifenyddes changes the pronunciation of the second ⟨y⟩. This is because the pronunciation of ⟨y⟩ depends on whether or not it is in the final syllable.

Stress on penultimate syllables is characterised by a low pitch, which is followed by a high pitch on the (unstressed) word-final syllable. In words where stress is on the final syllable, that syllable also bears the high pitch.^[13] This high pitch is a remnant of the high-pitched word-final stress of early Old Welsh (derived from original penultimate stress in Common Brittonic by the loss of final syllables); the stress shift from final to penultimate occurred in the Old Welsh period without affecting the overall pitch of the word.^[12]