mirror of
https://github.com/rhasspy/espeak-ng.git
synced 2026-10-31 13:38:47 +00:00
495 lines
8.7 KiB
Plaintext
495 lines
8.7 KiB
Plaintext
phoneme a
|
||
vwl starttype #a endtype #a
|
||
length 130
|
||
FMT(vwl_lv/a, 100) // adjusted from aa_3
|
||
endphoneme
|
||
|
||
phoneme a:
|
||
vwl starttype #a endtype #a
|
||
length 310
|
||
lng
|
||
FMT(vwl_lv/aa, 90)
|
||
endphoneme
|
||
|
||
phoneme E
|
||
vwl starttype #e endtype #e
|
||
length 120
|
||
FMT(vwl_lv/e2, 100)
|
||
endphoneme
|
||
|
||
phoneme E:
|
||
vwl starttype #e endtype #e
|
||
length 300
|
||
FMT(vwl_lv/ee2, 100)
|
||
endphoneme
|
||
|
||
phoneme e
|
||
vwl starttype #e endtype #e
|
||
length 120
|
||
FMT(vwl_lv/e, 100) // other alternative e_5
|
||
endphoneme
|
||
|
||
phoneme e:
|
||
vwl starttype #e endtype #e
|
||
length 300
|
||
FMT(vwl_lv/ee, 100) // other alternative e_8
|
||
endphoneme
|
||
|
||
phoneme i
|
||
vwl starttype #i endtype #i
|
||
length 120
|
||
FMT(vwl_lv/i, 110) // was i_6
|
||
endphoneme
|
||
|
||
phoneme i:
|
||
vwl starttype #i endtype #i
|
||
length 280
|
||
FMT(vwl_lv/ii, 90) // was i_7
|
||
endphoneme
|
||
|
||
|
||
phoneme y // English y, short Latgalian "hard i"
|
||
vwl starttype #i endtype #i
|
||
length 120
|
||
FMT(vwl_lv/y)
|
||
endphoneme
|
||
|
||
phoneme y: // Long Latgalian "hard i"
|
||
vwl starttype #i endtype #i
|
||
length 300
|
||
FMT(vwl_lv/yy)
|
||
endphoneme
|
||
|
||
phoneme o
|
||
vwl starttype #o endtype #o
|
||
length 120
|
||
FMT(vwl_lv/o, 90)
|
||
endphoneme
|
||
|
||
phoneme o:
|
||
vwl starttype #o endtype #o
|
||
length 300
|
||
FMT(vwl_lv/oo, 75)
|
||
endphoneme
|
||
|
||
phoneme u
|
||
vwl starttype #u endtype #u
|
||
length 120
|
||
FMT(vwl_lv/u, 140)
|
||
endphoneme
|
||
|
||
phoneme u:
|
||
vwl starttype #u endtype #u
|
||
length 300
|
||
FMT(vwl_lv/uu, 120)
|
||
endphoneme
|
||
|
||
phoneme ai
|
||
vwl starttype #a endtype #i
|
||
length 280
|
||
lng
|
||
FMT(vdiph/aai, 100)
|
||
endphoneme
|
||
|
||
phoneme au
|
||
vwl starttype #a endtype #u
|
||
length 300
|
||
lng
|
||
FMT(vdiph/aau, 100)
|
||
endphoneme
|
||
|
||
phoneme ei
|
||
vwl starttype #e endtype #i
|
||
length 280
|
||
lng
|
||
FMT(vdiph/eei_2, 100)
|
||
endphoneme
|
||
|
||
phoneme ie
|
||
vwl starttype #i endtype #e
|
||
length 250
|
||
lng
|
||
FMT(vdiph2/ie_2, 100)
|
||
endphoneme
|
||
|
||
phoneme iu
|
||
vwl starttype #i endtype #u
|
||
length 250
|
||
lng
|
||
FMT(vdiph2/iu_2, 100)
|
||
endphoneme
|
||
|
||
phoneme ui
|
||
vwl starttype #u endtype #i
|
||
length 250
|
||
lng
|
||
FMT(vdiph/ui, 100)
|
||
endphoneme
|
||
|
||
phoneme uo
|
||
vwl starttype #u endtype #a
|
||
length 270
|
||
lng
|
||
FMT(vdiph2/ua, 100)
|
||
endphoneme
|
||
|
||
phoneme uo` // shorter version of uo
|
||
vwl starttype #u endtype #a
|
||
length 150
|
||
lng
|
||
FMT(vdiph2/ua, 100)
|
||
endphoneme
|
||
|
||
phoneme oi
|
||
vwl starttype #o endtype #i
|
||
length 250
|
||
lng
|
||
FMT(vdiph/ooi, 100)
|
||
endphoneme
|
||
|
||
phoneme ts
|
||
vls alv afr nolink
|
||
voicingswitch J
|
||
length 55
|
||
lengthmod 1
|
||
WAV(ustop/ts2, 60)
|
||
endphoneme
|
||
|
||
phoneme tS
|
||
vls pla afr sib
|
||
voicingswitch dZ
|
||
lengthmod 2
|
||
length 80
|
||
WAV(ustop/tsh,40)
|
||
endphoneme
|
||
|
||
phoneme D // louder and shorter version of d
|
||
vcd alv stp
|
||
voicingswitch t
|
||
length 10
|
||
lengthmod 5
|
||
Vowelin f1=1 f2=1700 -300 300 f3=-100 80
|
||
Vowelout f1=2 f2=1700 -300 300 f3=-100 80 brk
|
||
|
||
IF PreVoicing THEN
|
||
FMT(d/xd, 110)
|
||
ENDIF
|
||
|
||
IF nextPh(isPause2) THEN
|
||
FMT(d/d_, 120) addWav(x/d_,110)
|
||
ENDIF
|
||
|
||
FMT(d/d, 110) addWav(x/d,110)
|
||
endphoneme
|
||
|
||
phoneme h // more quiet version of standard h
|
||
vls glt apr
|
||
lengthmod 3
|
||
ipa h
|
||
|
||
IF nextPh(#a) THEN
|
||
WAV(h/ha, 20)
|
||
ELIF nextPh(#e) THEN
|
||
WAV(h/he, 20)
|
||
ELIF nextPh(#i) THEN
|
||
WAV(h/hi, 20)
|
||
ELIF nextPh(#o) THEN
|
||
WAV(h/ho, 20)
|
||
ELIF nextPh(#u) THEN
|
||
WAV(h/hu, 20)
|
||
ELIF nextPh(isPause) THEN
|
||
WAV(h/h_, 40) // louder for foreign words
|
||
ENDIF
|
||
WAV(h/h_, 20) // no vowel following
|
||
endphoneme
|
||
|
||
phoneme j // **y**es
|
||
liquid pal starttype #i
|
||
|
||
IF nextPhW(isVowel) THEN
|
||
NextVowelStarts
|
||
VowelStart(j/j@)
|
||
VowelStart(j/ja)
|
||
VowelStart(j/je)
|
||
VowelStart(j/ji)
|
||
VowelStart(j/jo)
|
||
VowelStart(j/ju)
|
||
EndSwitch
|
||
VowelEnding(j/xj, 70) // In difference with English, vowel is lengthened, not shortened
|
||
ELSE
|
||
Vowelout len=127 // no vowel follows
|
||
FMT(j/xj)
|
||
ENDIF
|
||
|
||
IF prevPh(isPause) THEN
|
||
FMT(j/_j)
|
||
ENDIF
|
||
endphoneme
|
||
|
||
phoneme k
|
||
vls vel stp
|
||
lengthmod 2
|
||
voicingswitch g
|
||
Vowelin f1=0 f2=2300 200 400 f3=-100 80
|
||
Vowelout f1=0 f2=2300 300 400 f3=-100 80 rms=20
|
||
IF nextPh(isPause2) THEN
|
||
WAV(ustop/k_)
|
||
ENDIF
|
||
IF nextPh(#i) OR nextPh(l^) THEN
|
||
WAV(ustop/ki)
|
||
ENDIF
|
||
IF nextPh(l) THEN
|
||
WAV(ustop/kl)
|
||
ENDIF
|
||
WAV(ustop/k_unasp, 60) // weaker
|
||
endphoneme
|
||
|
||
phoneme c
|
||
vls pal stp pzd
|
||
voicingswitch J
|
||
WAV(ustop/c, 80)
|
||
endphoneme
|
||
|
||
phoneme l
|
||
liquid
|
||
|
||
// IF nextPh(isNotVowel) THEN
|
||
// ChangePhoneme(l/)
|
||
// ENDIF
|
||
|
||
IF nextPh(isNotVowel) THEN
|
||
PrevVowelEndings
|
||
VowelEnding(l/L1_@L)
|
||
VowelEnding(l/L1_aL)
|
||
VowelEnding(l/L1_eL)
|
||
VowelEnding(l/L1_iL)
|
||
VowelEnding(l/L1_oL)
|
||
VowelEnding(l/L1_uL)
|
||
EndSwitch
|
||
|
||
FMT(l/l_)
|
||
RETURN
|
||
ENDIF
|
||
|
||
CALL vowelstart_l
|
||
|
||
IF prevPh(a) OR prevPh(a:) THEN
|
||
VowelEnding(l/xl, 20) // FIXME
|
||
ELSEIF prevPh(isVowel) THEN
|
||
VowelEnding(l/xl, 10)
|
||
ENDIF
|
||
|
||
IF prevPh(isPause) THEN
|
||
FMT(l/_l)
|
||
ENDIF
|
||
|
||
IF prevPh(t) THEN
|
||
FMT(l/tl)
|
||
ENDIF
|
||
|
||
IF prevPh(l/) THEN
|
||
FMT(l/l_long) // double l, make it longer
|
||
ENDIF
|
||
|
||
FMT(l/l)
|
||
endphoneme
|
||
|
||
phoneme l^ // Latvian ļ, customized from standard l^
|
||
liquid
|
||
ipa ʎ
|
||
lengthmod 7
|
||
Vowelout len=60 lenadd rate
|
||
length 100
|
||
|
||
NextVowelStarts
|
||
VowelStart(l^/j2@)
|
||
VowelStart(l^/j2a)
|
||
VowelStart(l^/j2e)
|
||
VowelStart(l^/j2i)
|
||
VowelStart(l^/j2o)
|
||
VowelStart(l^/j2u)
|
||
EndSwitch
|
||
|
||
IF prevPh(isPause) THEN
|
||
FMT(l^/_l^)
|
||
ELSEIF nextPh(isPause) THEN
|
||
FMT(l^/l^_)
|
||
ENDIF
|
||
|
||
FMT(l^/l^)
|
||
|
||
endphoneme
|
||
|
||
|
||
|
||
|
||
phoneme m // copied from standard m phoneme with rule for 'mm'
|
||
vcd blb nas
|
||
Vowelout f1=2 f2=1000 -500 -350 f3=-200 80 brk
|
||
lengthmod 4
|
||
|
||
IF KlattSynth THEN
|
||
Vowelin f1=0 f2=1000 -50 -200 f3=-200 80
|
||
IF nextPh(isPause2) THEN
|
||
FMT(klatt/m_)
|
||
ENDIF
|
||
FMT(klatt/m)
|
||
ENDIF
|
||
|
||
NextVowelStarts
|
||
VowelStart(m/m@)
|
||
VowelStart(m/ma)
|
||
VowelStart(m/me)
|
||
VowelStart(m/mi)
|
||
VowelStart(m/mo)
|
||
VowelStart(m/mu)
|
||
EndSwitch
|
||
|
||
IF prevPh(isNotVowel) AND nextPhW(isLiquid) THEN
|
||
FMT(m/mj)
|
||
ELIF prevPh(isPause) THEN
|
||
FMT(m/_m)
|
||
ELIF nextPh(isNotVowel) THEN
|
||
FMT(m/m_)
|
||
ELIF nextPh(m) THEN // m/mj has no volume change, so doubled 'mm' merges better
|
||
FMT(m/mj)
|
||
ENDIF
|
||
endphoneme
|
||
|
||
phoneme n- // more voiced variant of n
|
||
vwl starttype #a endtype #a
|
||
ipa nU+0329
|
||
lengthmod 4
|
||
length 140
|
||
FMT(n/n-syl)
|
||
endphoneme
|
||
|
||
phoneme p
|
||
vls blb stp
|
||
lengthmod 2
|
||
voicingswitch b
|
||
Vowelin f1=0 f2=1000 -50 -100 f3=-200 80
|
||
Vowelout f1=0 f2=1000 -500 -350 f3=-300 80 rms=30
|
||
IF nextPh(isPause2) THEN
|
||
WAV(ustop/p_unasp_, 80)
|
||
ENDIF
|
||
IF nextPh(l) THEN
|
||
WAV(ustop/pl)
|
||
ENDIF
|
||
WAV(ustop/p_unasp, 120)
|
||
endphoneme
|
||
|
||
phoneme R // Latvian hard r
|
||
liquid trl
|
||
lengthmod 6
|
||
Vowelin f1=0 f2=1700 -300 300 f3=-300 80
|
||
Vowelout f1=2 f2=1700 -300 300 f3=-300 80 brk
|
||
FMT(r3/r_trill) addWav(r3/r_trill.wav, 60)
|
||
endphoneme
|
||
|
||
phoneme r // Latvian soft r
|
||
liquid trl
|
||
lengthmod 6
|
||
Vowelin f1=0 f2=1700 -300 300 f3=-300 80
|
||
Vowelout f1=2 f2=1700 -300 300 f3=-300 80 brk
|
||
FMT(j/j_) addWav(r3/r_trill.wav, 25)
|
||
endphoneme
|
||
|
||
phoneme s
|
||
vls alv frc sib
|
||
voicingswitch z
|
||
lengthmod 3
|
||
Vowelin f1=0 f2=1700 -300 300 f3=-100 80
|
||
Vowelout f1=0 f2=1700 -300 250 f3=-100 80 rms=20
|
||
WAV(ufric/s,60)
|
||
endphoneme
|
||
|
||
phoneme t
|
||
vls alv stp //lengthenstop
|
||
lengthmod 1
|
||
voicingswitch d
|
||
IF nextPh(isPause2) THEN
|
||
WAV(ustop/t_dnt, 40)
|
||
ENDIF
|
||
WAV(ustop/t_dnt, 50)
|
||
endphoneme
|
||
|
||
phoneme v // from v# in standard phonemes
|
||
vcd lbd frc
|
||
ipa ʋ
|
||
lengthmod 6
|
||
voicingswitch f
|
||
Vowelin f1=0 f2=1000 -300 -200 f3=-300 100
|
||
Vowelout f1=0 f2=1000 -500 -300 f3=-300 60 len=50
|
||
IF nextPh(isPause2) THEN
|
||
FMT(voc/v#_)
|
||
ENDIF
|
||
FMT(voc/v#)
|
||
endphoneme
|
||
|
||
phoneme v` // shortened v
|
||
vcd lbd frc
|
||
ipa ʋ
|
||
lengthmod 6
|
||
voicingswitch f
|
||
Vowelin f1=0 f2=1000 -300 -200 f3=-300 100
|
||
Vowelout f1=0 f2=1000 -500 -300 f3=-300 60 len=50
|
||
length 8
|
||
FMT(voc/v#_)
|
||
endphoneme
|
||
|
||
phoneme z` // Shortened version of z for diphthong dz
|
||
vcd alv frc sib
|
||
voicingswitch s
|
||
length 8
|
||
lengthmod 6
|
||
Vowelout f1=0 f2=1700 -300 300 f3=-100 60 len=50
|
||
|
||
IF KlattSynth THEN
|
||
Vowelin f1=0 f2=1390 -300 300 f3=0 0
|
||
Vowelout f1=1 f2=1390 -300 300 f3=-100 60 len=50 brk
|
||
ENDIF
|
||
|
||
IF nextPh(isPause2) THEN
|
||
length 20
|
||
FMT(voc/z_, 120) addWav(ufric/s_, 95)
|
||
ENDIF
|
||
FMT(voc/z, 110) addWav(ufric/s_, 100)
|
||
endphoneme
|
||
|
||
phoneme Z` // Shortened version of Z for diphthong dZ
|
||
vcd pla frc sib
|
||
voicingswitch S
|
||
length 30
|
||
lengthmod 6
|
||
Vowelin f1=0 f2=2000 0 300 f3=-200 80
|
||
Vowelout f1=2 f2=2000 0 300 f3=-200 80 brk
|
||
|
||
IF KlattSynth THEN
|
||
IF nextPh(isPause2) THEN
|
||
FMT(klatt/zh_, 70) addWav(vocw/zh, 70)
|
||
ENDIF
|
||
FMT(klatt/zh) addWav(vocw/zh, 130)
|
||
ENDIF
|
||
|
||
IF nextPh(isPause2) THEN
|
||
FMT(voc/zh_, 80) addWav(vocw/zh, 70)
|
||
ENDIF
|
||
FMT(voc/zh) addWav(vocw/zh, 130)
|
||
endphoneme
|
||
|
||
phoneme > // lengthen previous vowel less than using ":"
|
||
virtual
|
||
length 20 // it actually seems working shorter, but anyway...
|
||
endphoneme
|
||
|
||
phoneme } // Tiny, 5ms non-linked pause (used between phonemes in word)
|
||
pause
|
||
nolink
|
||
length 5
|
||
endphoneme
|
||
|
||
|
||
|
||
|