In Classical Quazian, which I am working on right now, the stop system has a six-POA /p t ʈ tʃ k q/ distinction, with voiced counterparts and prenasalized versions of both voiced and voiceless stops, giving 24 stop plosives. In addition to this, there are four nasals /m n ɲ ŋ/, four fricatives /s ʂ ʃ χ/ and a handful of sonorants /w l ɭ j ʀ/.

a. Prenasalized stops become something else, depending on dialect. Popular options are:
*ⁿT ⁿD -> D N
*ⁿT ⁿD -> N̥ N
*ⁿCV -> CṼ

or some combination of these.

b. A depalatalization occurs in most dialects, eliminating retroflexes and palatoalveolars from the system, as follows:
*t d s-> ts dz h -> s z h
*ʈ ɖ ʂ ɭ-> t d s ɾ
*tʃ dʒ ʃ -> ts dz s

c. Uvulars change to something else, depending on dialect:

*q -> ʔ (usually; sometimes remains in initial position)
*ɴ (from *ᶰɢ) -> ŋ (also almost everywhere; the prenasalized uvulars end up with the same reflexes as their velar counterparts.)

The fate of *ʀ may differ from dialect to dialect:

a) *χ and *s merge into /h/; *ʀ and *ɢ merge into /ɦ/.
*χ, *s -> h
*ʀ, *ɢ -> ɦ

b) Similar to a), but with uvulars fronting to velars and *χ having a different reflex than *s:
*χ -> x
*ʀ, *ɢ -> ɣ

c) Similar to b), but *ʀ pairs with *χ instead of with *ɢ:
*χ, *ʀ -> x
*ɢ -> ɣ

Situation c) has a few subvariants, including merging resulting /x/ with /h/ and/or merging /ɣ/ with /g/.

Somewhere along the way, an entirely unextraordinary change of *w to /v/ occurred.

A standard-looking phonetic inventory for consonants would look something like this:

p	t		k	ʔ
b	d		g
m	n	ɲ	ŋ
	s		x	h
	z		ɣ
v	ɾ	j
	l
	ts
	dz

Some common variants of this inventory are: adding a set of voiceless nasals; eliminating the velar fricatives and adding /ɦ/; retention of /q/. Some sample CQ words with their modern Quazian reflexes (note that codas generally underwent different changes):

*pɨ́t -> pɨ́ʔ
*ṣâl -> sâu
*ᵑgâɲ -> ŋẽ̂
*qân -> ʔɑ̃̂
*ʈúsa -> túhə
*ᵐpáʀ -> bɑ̂
*kúnayak -> kúnəyəʔ
*náq -> *nɑ́ʔ
*čətúk -> tsəsúʔ
*dôʀ -> zuò
*dʒáw -> dzó
*χît -> xìʔ
*ⁿʈéʀ -> dié

Thoughts? I'm most worried about my depalatalization changes.

Depalatalization looks fine.

Prenasalized stops: what did that one Bantu language do with them?

In Vian, I had voicing transfer to tone and prenasalized stops create new voice contrasts: V'eng ta da nta nda :> Vian tā tà dā dà.

Yeah, tone is another thing that varies depending on dialect. I've mostly avoided tying down a description of the language to any specific dialect, because Quazian is only beginning to experience dialect levelling, its writing system has a large number of logograms and is outdated even in its phonetic component. You can talk about the dialect of the capital city, but this isn't the "standard dialect" the way we think of them.

I'm actually planning on making an isogloss map for the language at some point...I'd need a country map first, though.

Hmm maybe I should generate tone in Kotic by merging voiced and voiceless consonants, with voicing conditioning low tone and voicelessness conditioning high tone, and then have the tonality come under the control of the nonconcatenativity...

I need to figure out what happens with the resulting tones after I collapse my obstruents to /p pʼ t tʼ ts tsʼ tʂ tʂʼ tɕ tɕʼ k kʼ q qʼ ɸ θ s ʂ ɕ x χ/. Obviously the resulting tones will not be attached to the vowels in the roots, because the vowels in the roots are easily dislocated from their original positions. I first will need a notation for marking tone in roots, as tone will be associated with the consonants and not the vowels, maybe I will continue marking down roots as if they had voiced consonants just as a notation for marking tone (old voiced consonant = low tone, old voiceless consonant = high tone).

You could create sound changes by having some of the tones move with the vowels. I'm thinking more non-systematic than "tone 1 stays, tone 2 moves with vowel, no exceptions", to be clear.

Does that happen? That sounds pretty spooky. Why are tones so spooky? (And should I really be asking this at 7 in the morning?)

the silhouette of a naked woman!

never has such wickedness been shown with so little shame

…missed that part.

but that's the whole thing

Not the sound! I saw it on the thumbnail yeah but after that I forɡot it was there because I tabbed back here to paste in the video code (or hash or whatever it's called).

My proto-language has a dorsal R, which is probably closer to /ʕ/ and in any case is a rip-off of *h₂. This gets lost pretty much everywhere but word-initially and after a consonant. I also have /d/ > /z/ > /ɹ/ intervocalically (/d/ > /z/ in other environments, but elsewhere it's just lost.) This gives me /ɹ ʁ/, but I only have one rhotic in my orthography (and I want it that way.) Should I merge the two, or keep them separate but with identical writing? I admittedly could change one to a different sound, but I'd really like to have a rhotic in both of those environments.

most languages with voiced dorsal fricatives/approximants don't pattern them as rhotics, 'rhotic' is only a natural class in the sense that /r/ does weird things but people who speak a language where /r/ does those things tend to have an intuition that they're forms of /r/, like the velar lateral affricate in Hiw