[New post] Experimenting With Stop Consonant Transitions
synsinger posted: " I've got two mutually exclusive features I'd like synSinger to have. I want the reclist ("recording list") to be sufficiently small that it won't take forever to record and mark up the recordings needed to create a list. But I also want the quality"
I've got two mutually exclusive features I'd like synSinger to have.
I want the reclist ("recording list") to be sufficiently small that it won't take forever to record and mark up the recordings needed to create a list.
But I also want the quality of the output to be sufficiently good.
My current approach is to record all the CV and VC combinations (vowel/consonant transitions). That gives reasonably good coverage.
I'll also include the most common blends, such as /B L/ and /S T R/. However, I'd like to avoid having to record every possible combination.
So when it comes to CC transitions (consonant/consonant) transitions,I'm trying to find any replacements I can use to avoid recording transitions that might not be necessary.
Stop consonants are so-called because the sound "stops" before the consonant. For example, here's /AA D AA/:
The spoken phonemes /AA D AA/
If you look at the above sound, you can see that the sound doesn't really stop before the /D/ sound. In fact, there's a constant voicebar throughout.
However, if the "stop" is replaced with an actual stop, there's virtually no audible difference:
The same recording, with silence at the "stop"
Now, there are sounds the are very important here. For example, if the /D/ were replaced with one that targeted a different vowel, the might not sound like a /D/ at all. At the length of the stop also helps indicate what the sound is supposed to be.
But in general, phonemes can be split at stop consonants without changing the perception of the sound.
But what about CC transitions, where the initialsound is a stop consonant?
Blends are special cases. Blends are formed by all the liquid consonants /L/, /R/, /W/ and /Y/. These all require individual recordings.
But where there isn't a blend, there's often an additional closure after the stop consonant. The following example is "Batman", /B AE T M AE N/:
"Batman"
In regular speech, the /T/ is replaced by a glottal closure, so perhaps this isn't the best example. Here's the nonsense syllable "akvia":
The nonsense syllable "akvia"
Once again, there's a stop following the stop consonant.
So the general rule appears to be that non-blending initial stop consonants in CC blends can be followed by stops,
So what happens with a CC transition when both consonants are stop consonants?
The answer is... it depends.
In some cases, such as "outdoor", if the /T/ is pronounced, it sounds odd:
The word "outdoor" with the /T/ articulated - not normal sounding.
What actually happens is that the /T/ doesn't get articulated, and instead gets subsumed into the /D/:
The word "outdoor" without the /T/ articulated
The stop is there for the /T/, there's just no plosive - it's basically a glottal stop.
This is similar to what happens in the the word "abdicate":
No comments:
Post a Comment