I've been slowly (emphasis on slow) working on my next bank which will be multi-pitch, multi-append Korean and decided to share the fruits of my labor with the community - the reclist. Hopefully anyone who uses this finds it to be comprehensive, useful, and a good update to Syo's list from 2012.
- All sounds necessary for modern Korean (standard Seoul dialect)
- Optional English that was included (or missing) from SeeU's intended English recordings (will not fully replace the functionality of a proper English voicebank, but will make UST editing easier due to the frequency of English in commercial K-pop works)
- Pronunciation guide including references to Japanese, English examples (General American/Received Pronunciation), IPA, and X-Sampa pronunciation.
*I prefer to record single recordings where possible and simply combine them into a single wav file. How you record this bank depends on your recording software of choice and convenience.
- Why do I need Korean/Hangul in the pronunciation guide? -Anyone who owns SeeU (시유) or plans to get Uni (유니) will need to know how to read/type Korean to use them properly and will also need to know Korean for transliterating to romaja for UTAU. While many K-pop songs have romaja lyrics, not all of them are accurate or some will use McCune–Reischauer (North Korean romaja) over Revised Romanization (South Korean romaja - what we'll be using).
- Why the underscores/dashes in the reclist? -Refer to the Mora Count. They simply allow me to easily edit a txt file for single string recordings (for ex. a single wav file for consonants that don't have morphemes such as m, n, or s). The dashes are for easy reading for Vowel-Vowel, not glottal stops/pauses.
- Why the brackets next to certain sounds? -The brackets - assuming Seoul dialect is used (rather than differentiation for dialect/stylistic purpose) - is to let you know what sounds will require double/triple alias.
a ||아 = same pronunciation as Japanese あ
eo || 어 = same pronunciation as the u in words 'sun,' 'fun,' 'run.'
o || 오 = closed pronunciation, similar to the o sound in 'owe'
u || 우 = similar to the o sound in 'who'
eu ||으 = same pronunciation as japanese う
i || 이 = same pronunciation as japanese い
ae || 애 = in modern Seoul dialect, has the same pronunciation as 'e.' Traditionally pronounced with a higher palate, sounding a little closer to an @ sound (such as the word bat).
e || 에 = same pronunciation as japanese え
oe || 외 = sounds like the word 'wet' without the t. Also romanized as oi. Sounds like both we and wae in Seoul dialect.
wi || 위 = combination of 'u' and 'i' sounds like Japanese うぃ
we || 웨 = sounds like oe and wae in Seoul dialect
wae || 왜 = sounds like oe and we in Seoul dialect
weo || 워 = sounds like the word 'won' (or the Korean dollar ₩). sometimes romanized as wo.
ui || 의 = combination of 'eu' and 'i.' combined with a consonant, sounds like a nasalized 'i' sound. Transforms into an 'e' or 'ye' sound at the end of certain words - close listening and phoneme swaps will be necessary for UST creation. ex. 나의 (naui) na + ui or na + e (2nd pronunciation is informal and used more often in songs). Also romanized as eui.
ng || ㅇ = ieung. In Korean writing, serves as a stand-in consonant for words that phonetically start with a vowel (proper written hangeul requires a consonant; vowels are never standalone). ex. 아이 (written as 'ai' in romaja) NOT ㅏㅣ, 아ㅣ, or ㅏ이. Secondary position transforms this into an 'ng' sound.
ㄱ = giyeok. romanized as g and k. initial pronunciation (beginning of words) is unaspirated k. pronunciation in the middle of words/sentences is an unaspirated g. ending consonant is pronounced as an unaspirated k.
ㄴ = nieun. sounds like 'n' in all situations.
ㄷ = digeut. romanized as d and t. initial pronunciation (beginning of words) is unaspirated t. pronunciation in the middle of words/sentences is an unaspirated d. ending consonant is pronounced as an unaspirated t.
ㄹ = rieul. romanized as r, ll, and l. initial pronunciation (beginning of words) is 4 (aveolar tap). pronunciation in the middle of words/sentences - as well as the ending consonant - is l.
ㅁ = mieum. romanized as m. sounds like m in all situations.
ㅂ = bieup. romanized as p and b. initial pronunciation (beginning of words) is unaspirated p. pronunciation in the middle of words/sentences is an unaspirated b. ending consonant is pronounced as an unaspirated p.
ㅅ = siot. romanized as s and t. pronunciation is a soft s (mouth must be in the position to make an s sound, but teeth must not be touching). combined with 'i' or 'y,' pronunciation becomes a soft sh (like in 's,' teeth must not touch). when written in as a final consonant, 's' sound is replaced with 't.'
ㅈ = jieut. romanized as j. initial pronunciation is a soft ch (tongue must be in the position to make a ch sound, but teeth must not be touching). in the middle of words, pronunciation is a soft j. when written in as a final consonant, 'ch' sound is replaced with 't.'
ㅊ = chieut. romanized as ch. pronunciation is an aspirated ch sound. when written in as a final consonant, 'ch' sound is replaced with 't.'
ㅋ = kieuk. romanized as k (k' or kh for the sake of utau alias). pronounced as an aspirated k. when written in as a final consonant, phonetically replaced by giyeok (unaspirated k).
ㅌ = tieut. romanized as t (t' or th for the sake of utau alias). pronounced as an aspirated t. when written in as a final consonant, phonetically replaced by digeut (unaspirated t).
ㅍ = pieup. romanized as p (p' or ph for the sake of utau alias). pronounced as an aspirated p. when written in as a final consonant, phonetically replaced by bieup (unaspirated p).
ㅎ = hieut. romanized as h. pronounced as h. when written as a final consonant, replaced by digeut (unaspirated t). if it comes after a patchim (final consonant) in the middle of a word, the preceeding consonant becomes aspirated. ex. 못해 (mothae = mo-t'ae)
ㄲ = ssang giyeok (kk or gg). a sharp, forced unaspirated k. can be double aliased with initial ㄱ
ㅆ = ssang siot (ss). a stronger 's' sound (without teeth meeting).
ㄸ = ssang digeut (tt or dd). a sharp, forced unaspirated d. can be double aliased with secondary ㄷ
ㅃ = ssang bieup (pp or bb) a sharp, forced unaspirated b. can be double aliased with secondary ㅂ
ㅉ = ssang jieut. a stronger j sound, similar to English. For ex. judge or jeep
v = not native to Korean, but included in SeeU's hidden phonemes.
f = not native to Korean, but included in SeeU's hidden phonemes.
R = English R. not native to Korean or included in SeeU's hidden phonemes. useful for Konglish.
z = English z. not native to Korean, but included in SeeU's hidden phonemes.
c = English 's.' not native to korean, but included in SeeU's hidden phonemes. this list expands it's use beyond SeeU's capabilities
@r = ending 'er' sound such as the word 'singer.' not native to korean but included in SeeU hidden phonemes.
-D = english 'th' sound such as 'this.' not native or korean or included in SeeU's hidden phonemes, but added for additional English capability.
*** Standalone vowels and glides
*** Secondary ㅇ, ending consonant
*** Vowel-Vowel. Optional for a shorter recording session, but generally recommended
*** Initial-Secondary-Ending consonant (record final CVs shortly or the last [hard] consonant as a glottal stop depending on your preference).
hoe_ [hwe, hwae]
*** kk/gg, tt/dd, and pp/bb may be double aliased using prior recordings for convenience; only ss and jj are truly phonetically different from s and j.
*** These aren't necessary for pure Korean, but is meant to expand capabilities where Konglish is concerned (English is used a lot in K-pop music). SeeU was intended to be tri-lingual with Korean, Japanese, and English. Uni will presumably release Korean this fall/winter and English next summer.
What a bank may sound like using this reclist:
NOTE: I am not Korean in ethnicity or nationality, nor do I speak it as a first language. I'm merely a lover of the language and South Korean media and am in the beginner's stages of learning it.
NOTE 2: With the least amount of recordings (no differentiation between e/ae/we/wae/oe/ye/yae, skipping kk, tt, pp, skipping VV, and omitting all extras) the amount needed for Korean is potentially 215 samples. The full bank as currently presented (for the over achieving lot) is 484 samples.
LATEST EDITS MADE:
- addition of wi in pronunciation guide
- correction of oe's hangeul equivalent