mirror of
https://github.com/rhasspy/espeak-ng.git
synced 2026-10-31 05:35:35 +00:00
340 lines
5.9 KiB
Plaintext
340 lines
5.9 KiB
Plaintext
phoneme a
|
||
vwl starttype #a endtype #a
|
||
length 130
|
||
FMT(vwl_lv/a, 100) // adjusted from aa_3
|
||
endphoneme
|
||
|
||
phoneme e
|
||
vwl starttype #e endtype #e
|
||
length 120
|
||
FMT(vwl_lv/e, 100) // other alternative e_5
|
||
endphoneme
|
||
|
||
phoneme i
|
||
vwl starttype #i endtype #i
|
||
length 120
|
||
FMT(vwl_lv/i, 110) // was i_6
|
||
endphoneme
|
||
|
||
phoneme o
|
||
vwl starttype #o endtype #o
|
||
length 120
|
||
FMT(vwl_lv/o, 90)
|
||
endphoneme
|
||
|
||
phoneme u
|
||
vwl starttype #u endtype #u
|
||
length 120
|
||
FMT(vwl_lv/u, 100)
|
||
endphoneme
|
||
|
||
phoneme ai
|
||
vwl starttype #a endtype #i
|
||
length 280
|
||
lng
|
||
FMT(vwl_lv/ai, 100)
|
||
endphoneme
|
||
|
||
phoneme au
|
||
vwl starttype #a endtype #u
|
||
length 280
|
||
lng
|
||
FMT(vwl_lv/au, 100)
|
||
endphoneme
|
||
|
||
phoneme ei
|
||
vwl starttype #e endtype #i
|
||
length 280
|
||
lng
|
||
FMT(vwl_lv/ei, 100)
|
||
endphoneme
|
||
|
||
phoneme ie
|
||
vwl starttype #i endtype #e
|
||
length 250
|
||
lng
|
||
FMT(vwl_lv/ie, 100)
|
||
endphoneme
|
||
|
||
phoneme iu
|
||
vwl starttype #i endtype #u
|
||
length 250
|
||
lng
|
||
FMT(vwl_lv/iu, 100)
|
||
endphoneme
|
||
|
||
phoneme ui
|
||
vwl starttype #u endtype #i
|
||
length 250
|
||
lng
|
||
FMT(vwl_lv/ui, 100)
|
||
endphoneme
|
||
|
||
phoneme uo
|
||
vwl starttype #u endtype #a
|
||
length 280
|
||
lng
|
||
FMT(vwl_lv/ua, 110)
|
||
endphoneme
|
||
|
||
phoneme oi
|
||
vwl starttype #o endtype #i
|
||
length 250
|
||
lng
|
||
FMT(vwl_lv/oi, 100)
|
||
endphoneme
|
||
|
||
phoneme ts
|
||
vls alv afr nolink
|
||
ipa t͡s
|
||
voicingswitch J
|
||
length 55
|
||
lengthmod 1
|
||
WAV(ustop/ts2, 60)
|
||
endphoneme
|
||
|
||
phoneme tS
|
||
vls pla afr sib
|
||
ipa t͡ʃ
|
||
voicingswitch dZ
|
||
lengthmod 2
|
||
length 80
|
||
WAV(ustop/tsh,40)
|
||
endphoneme
|
||
|
||
phoneme D // louder and shorter version of d
|
||
vcd alv stp
|
||
ipa d
|
||
voicingswitch t
|
||
length 8
|
||
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 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, 127) // 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
|
||
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(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 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 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
|
||
liquid trl
|
||
ipa r
|
||
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 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 gH // arabic Ghain, the phoneme is derived from Q
|
||
vcd vel frc
|
||
voicingswitch R2
|
||
ipa ɣ
|
||
Vowelin f1=2 f2=2300 200 300 f3=-300 80
|
||
Vowelout f1=2 f2=2300 250 300 f3=-300 80 brk
|
||
FMT(g/g) addWav(vwl_ar/dhad)
|
||
endphoneme
|
||
|
||
phoneme dH //arabic Dhad TODO this probably should be simple d from standard phonemes
|
||
vcd alv stp
|
||
ipa dˤ
|
||
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)
|
||
ENDIF
|
||
IF nextPh(isPause2) THEN
|
||
FMT(d/d_) addWav(x/d_)
|
||
ENDIF
|
||
FMT(d/d) addWav(x/d)
|
||
endphoneme
|
||
|
||
phoneme s[
|
||
vls dnt stp
|
||
voicingswitch s
|
||
IF nextPh(s[) THEN
|
||
length 50
|
||
ELSE
|
||
length 160
|
||
ENDIF
|
||
lengthmod 7
|
||
WAV(vwl_ar/shin)
|
||
endphoneme
|
||
|
||
phoneme X
|
||
vls uvl frc
|
||
voicingswitch Q
|
||
lengthmod 7
|
||
Vowelin f1=0 f2=2300 200 400 f3=-100 80
|
||
Vowelout f1=0 f2=2300 300 400 f3=-100 80 rms=20
|
||
WAV(ufric/xx, 70)
|
||
endphoneme
|
||
|
||
|