☢ Recording ☢
☆ Core Reclist ☆
CVVC
CVVC strings are preceeded by the prefix [_+_]. They should be recorded in continuous strings with no breaks.
The strings are organized by vowel to help with consistant pronunciation. Consonants within strings are organized by first by manner of articulation, then place of articulation, then voicing.
Most strings are written in a pattern like _+_V-CV-CV-CV..., though there are a few exceptions to help with pronunciation, and some strings use the dummy vowel [@] if the CV of a particular combination is needed but the VC is not.
Nasal vowel CVVC strings are organized like _+_@-CVN-@-CVN... with a dummy vowel and dummy coda to make pronunciation easier.
The FULL list also has some syllables starting with [s], like spV-st@-sk@; these are where the unaspirated allophones of those plosives will be sampled from.
Vowels
There are two types of vowel strings: single-vowels and vowel blends. Both are preceeded by the prefix [___].
The single-vowel strings are organized like ___V'V-V for monophthongs and ___V'VI-V-VI for dipthongs. The ['] represents a glottal stop, and the [V-V] represents a sustained vowel.
For dipthongs and R/L-Coloured vowels, the first portion of the vowel is what should be extended, as indicated by the way its written. Syllabic consonants, including [@r] and[@l], should be treated the same as monophthongs.
The vowel blends should be recorded like those in a VCV reclist, alternating between each vowel continuously.
These strings are organized into tense-tense blends and tense-lax blends. Lax-tense and lax-lax pairs are not needed because lax vowels do not occur in open syllables — though the phonemically lax [@], [{], [@r], and [@l] are exceptions to this rule, and therefore included in the tense position.
Vowel blends are also not included for R/L-Coloured vowels because it made this section impractically long, and these can be synthesized through other sample combinations.
Consonants
Like vowels, there are two string types in this section: intial/final consonants and consonant blends. The former is prefixed with [+__] and the latter by [C__], where C is the coda of the syllables.
The initial/final consonant strings consist of samples meant to be blended with CVs and VCs, respectively. This is to alieviate the need for initial CVs and final VCs, significantly reducing the size of the reclist. They should be recorded with a beat of silence between each syllable.
The coda-to-onset consonant blends are the largest group of strings in either reclist, but also the most versatile, as these samples can be used both for blending the coda of one syllable into the onset of the next, as well as constructing consonant clusters.
Most strings are organized like C1__@C1-C2@C1-C3@C1-C4@C1..., where C1 acts as the coda of each syllable.
Some strings are split like C1__@C1-C2@__@C1-C3@, with one or two beats of silence in the middle, if the blends being sampled are particularly tricky to pronounce. This makes the section a bit longer, but saves time and frustration by making recording easier.
Initial plosives and plosive-plosive blends are excluded, because the result would just be silence (as explained on the configuration page).
☆ Optional Add-ons ☆
Consonant Clusters
The consonant cluster strings are quite similar to the normal consonant blends, but can provide more naturalistic production. They are preceeded by the prefix [++_].
The medial onset clusters are similar to CVVC strings, written in a pattern like ++_@-CC@-CC@-CC@....
Medial coda clusters are only needed if they have three or more consonants, as the final consonant of the cluster becomes the onset of the second syllable, so clusters with only two consonants would be identical to the normal blends. They use the split string style like ++_@CC-C@__@CC-C@ for easier pronunciation.
Initial onset and final coda cluster strings are handled like the initial/final consonants, with beats of silence between each syllable. They're written like ++_CC@_CC@_CC@... and ++_@CC_@CC_@CC....
Much like everything else in the reclist, all clusters are written phonemically. This means clusters with inserted / deleted consonants are not transcribed as such, even if they are quite common, such as an inserted [p] in the cluster [ms].
Breaths
This add-on is very short, and pretty straightfoward. Strings are prefixed with [+++_] and consist of two syllables, each either a vowel or a dummy vowel with a coda, followed by [hh]. The [hh] represents an exhale at the end of the utterance.
CVC
To be up front, I really don't find these samples necessary for inclusion, and debated cutting them out from the list entirely. They add over 100 strings and over 700 oto lines just to the LITE list, for a lot of samples that will rarely be used and don't add that much to the overall quality. The FULL list gains over 200 strings and over 1100 oto lines.
That being said, for a something like a monopitch voicebank prioritizing naturalistic utterances, I can see why they would be beneficial, so I've written out this add-on anyways.
These strings, like the CVVCs, are organized by vowel, but like the initial/final consonant strings, each syllable is separaved by a space, like +__CVC_CVC_CVC.... They share the same prefix as the initial/final strings, [+__], because if you are recording this add-on, you do not need to include both string types.