Hey !
This seems almost perfect. After adding in your code to where '</v>' appears in my version of SuperTextMesh.cs (line 1870), voices close out properly, making no additional sound as well as without closing out the '<transcribe>' effect with it. I have not tested a voice and colour combination at this time to gauge whether '</v>' would reset the colour, though it doesn't entirely appear so as the glyphs are all coloured as they should be. So that's a success !
However, the effect with '<transcribe>' being escaped is still present. Here is what the line currently looks like:
<w><transcribe>Lorem ipsum</transcribe><c=fire><transcribe> dolor sit</transcribe></c><transcribe> amet</transcribe>, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua.</w>
I have noticed that if I put '<c=fire>' within the next opening '<transcribe>' tag (now reading "[..]<transcribe>Lorem ipsum</transcribe><transcribe> <c=fire>dolor sit</transcribe></c>[..], the '<transcribe>' effect is still escaped with 'dolor sit <transcribe>' now having the fire effect. This means the pre-parser only checks for the first instance of '<transcribe>' and '</transcribe>' without amending the other calls.
Otherwise, without an intermediary close of '<transcribe>' and merely putting '<c=fire>' and '</c>' around some other glyphs, they result in the expected behaviour (which is essentiallyno effect since the '<transcription>' effect overrides it with its own colour definition as expected).
While I can certainly work around this for properly-emphasised effects (and properly-written dialogue), it does make me wonder if there'd be some way to override the next colour definition (perhaps by adding an attribute like "overrideNext=true" to a colour ? e.g. <c=fire overrideNext=true>). However, if STM effects like colours and voices allow spaces in their names instead of using them as a separator, this would become problematic. Just a thought problem !
I think with this, I am almost fully satisfied with the work; it would just be the multiple '<transcribe>' calls ability that would bring it all together. Thank you so much for all your help !