Does not work for me.
Could you please provide an example how to run the model to get the Belarusian speech output?
All examples from microsoft/speecht5_tts work for me, but when I switch the model to use KoRiF/speecht5_finetuned_common_voice_be and provide a Belarusian text in UTF-8 then I hear nothing but noise.
Here is the original model vocab, without some transliteration or a full retrain with a vocab including cyrillic letters it cannot possibly work :-(
Did you use some custom text normalization function to map Cyrillic into Latrin script?
0 0
0
0
▁ -1.70962
e -2.31708
t -2.6521
a -2.80277
o -2.83474
n -2.95012
i -3.00878
h -3.0161
s -3.04168
r -3.10175
d -3.38156
l -3.46726
u -3.81492
c -3.98254
m -3.98623
f -4.05427
w -4.05972
g -4.18846
y -4.20274
, -4.30063
p -4.32395
b -4.52745
. -4.65942
v -4.94521
k -5.07639
" -5.25082
I -5.56291
' -5.84526
T -5.93671
A -6.3936
S -6.55078
H -6.56163
; -6.70453
x -6.72004
W -6.76669
- -6.79379
B -6.80165
? -6.99726
C -7.0171
M -7.05023
! -7.16498
q -7.17415
j -7.18821
E -7.24076
N -7.27363
P -7.29694
O -7.31707
D -7.42123
L -7.44959
G -7.54578
R -7.55282
F -7.57062
Y -7.67737
z -7.78446
J -8.11335
: -8.18033
K -8.57823
U -8.74393
V -8.86835
) -9.54172
( -9.58167
Q -9.93155
Z -10.7209
] -11.832
[ -11.9003
X -12.0068
— -12.2907
/ -12.8318
æ -15.1091
é -15.8023
{ -16.4954
} -16.4954
ê -16.4954
œ -16.4954
̄ -16.4954