It seems native English SynthV voicebanks tend to express American accent, and so are non-English voicebanks in cross-language synthesis. Though, in case that the developers are willing to support British English as well, here are my suggestions.
Though ultimately, fully supporting British English would require adding it as a separate “language”, I think adding only a dictionary for it can act as a temporary solution.
The most notable difference between American English and British English is the pronunciation of vowels. To support British English, I suggest adding two phonemes, namely the vowel o
for [o], and the semivowel ex
for [ə̯].
That summarizes the phonemes of the English keywords as follows:
Keyword | American phonemes (current SynthV dict) |
British phonemes |
---|---|---|
TRAP | tr ae p |
tr ae p |
BATH | b ae th |
b aa th |
PALM | p aa l m |
p aa l m |
LOT | l aa t |
l ao t |
CLOTH | k l ao th |
k l ao th |
THOUGHT | th ao t |
th o t |
KIT | k ih t |
k ih t |
DRESS | dr eh s |
dr eh s |
STRUT | s tr ah t |
s tr ah t |
FOOT | f uh t |
f uh t |
FACE | f ey s |
f ey s |
GOAT | g ow t |
g ow t |
FLEECE | f l iy s |
f l iy s |
GOOSE | g uw s |
g uw s |
PRICE | p r ay s |
p r ay s |
CHOICE | ch oy s |
ch oy s |
MOUTH | m aw th |
m aw th |
NURSE | n er s |
n ax s |
START | s t aa r t |
s t aa t |
NORTH | n ao r th |
n o th |
FORCE | f ao r s |
f o s |
NEAR | n ih r |
n ih ex |
SQUARE | s k w eh r |
s k w eh |
CURE | k y uh r |
k y uh ex |
COMMA | k aa m ah |
k ao m ax |
LETTER | l eh dx er |
l eh t ax |
HAPPY | hh ae p iy |
hh ae p iy |
Hope this is motivating enough.