DISREGARD THAT, I SUCK COCKS
In the earliest written records of Rau*, the consonant inventory is superficially similar to that of Old Kangshi, the source of the Rau script. All Old Kangshi consonants except /z r/ are also attested in Rau, and two consonant series were innovated by alteration of the series of /s d ts dz/, giving the appearance of a phoneme inventory of: (writing ŋ ʁ instead of the ñ h of Kangshi romanization)
m n ŋ
p t ts ts₂ k q
b d d₂ dz dz₂
s s₂ x
d2 = l
s s2 = sh lh
The Old Rau attestations n.baxdzun and n.ts₂twxwb for the place-names Bâghzon and Ltoúhoub demonstrate that x was used for both the velar and the uvular fricatives, but this is the only defect in the script. These two sounds are in near-complementary distribution, with /χ/ occurring in back-vocalic environments and /x/ occurring otherwise, but diphthong loss and syncope predating Old Rau created some unpredictable occurrences - cf. the Ngabtou name for Baghzon, /baixəɟuːŋ/.
This near-complementary distribution of velars and uvulars was further disturbed by the merger of ts₂ dz₂, identifiable as lateral obstruents by vacillations such as the aforementioned n.ts₂twxwp for Ltouhoub and by comparison with other Hathic languages, into k g. However, a robust contrast between /g/ and /ʁ/ must be reconstructed for the common ancestor of Rau and Ngabtou, both of which reflect Continental Hathic *r as a dorsal, although Ngabtou shows secondary fronting - the Ngabtou culture hero /paɾaɣǐn-tuɣàn/ is known in Rau as Houlrỉn. (The shift of *p to Rau /χ/ is irregular but not uncommon.)
As for the other consonants, /g/ was a plosive (rather than a fricative, as it is in Kangshi) and s s₂ d₂ correspond to /s ɬ l/. Precise phonetic values for the affricates ts dz cannot be established; they correspond to Proto-Continental Hathic *c *ɟ, were presumably affricates when Rau began to be written, and are reflected in the spoken Rau of the current year as /θ ɹ/.
The consonants of spoken Rau in the current year are, in an abstract analysis that largely follows the romanization:
/m n ŋ/ <m n ng>
/p t (ʈ) k q/ <p t tt k q>
/b d ɹ g~j ʁ/ <b d z g r> (note: /g/ is [j] in codas)
(/θ͈ s͈ ɬ͈ x͈ χ͈/) <st sht lt ghk hq>
/θ s ɬ x χ/ <s sh l gh q>
+ unwritten /ɣ/ excreted after word-final /ɯ/ which may be unstressed
These consonants can be divided into fortis-lenis pairs. Nasals are lenis and unpaired, although one could just as well posit unit prenasalized stops /mb nd ŋg ɴɢ/ as the fortis variants of nasals. /ʈ/, the fortis analogue of /ɹ/, has no regular diachronic origin and was introduced by loans from western Vengic languages like Narng and Qoa; many speakers merge it into /t/. Fortis consonants are always unvoiced, articulated more forcefully, often aspirated, and prohibited in codas; lenis consonants are often voiced, tend to coalesce with unstressed schwa to produce syllabic consonants, and are allowed in codas. In fact, the series /b d ɹ g ʁ/ is preferentially syllabified into the coda (unless an onset slot of a stressed syllable is available) and exerts great effect on preceding vowels: shtíbor 'skin' is pronounced [ɕʰjúwɐʁ]. (Fortis fricatives developed from a sequence of fricative and homorganic fortis stop, although this is blocked below a certain level of juncture.)
It is late here and I don't know how vowels work yet so here is an example sentence that ignores tones.
in bzogs te mnar blqudou na harib hshtir boa toug ash beshtad.
[m.brojθ.tʰə mnɐβ.ɬqʰə.dɯɣ.nə χɐ.juwχ.ɕʰjɐw.ɐʁ tʰɨɣ.ʒ.β.ʃʰæð]
There are various juncture levels, one tone-bearing stressed vowel per Copticesque phrasal group, all the normal Rau things. I think the unstressed vowels are dominant ɐ and recessive schwa which is raised to barred i by velars incl. /j/ or something.
* Rau = Xaukatai = Kastas