<s>
Speech	B-Application
synthesis	I-Application
is	O
the	O
artificial	O
production	O
of	O
human	O
speech	O
.	O
</s>
<s>
A	O
computer	O
system	O
used	O
for	O
this	O
purpose	O
is	O
called	O
a	O
speech	B-Application
synthesizer	I-Application
,	O
and	O
can	O
be	O
implemented	O
in	O
software	O
or	O
hardware	B-Architecture
products	O
.	O
</s>
<s>
A	O
text-to-speech	B-Application
(	O
TTS	O
)	O
system	O
converts	O
normal	O
language	O
text	O
into	O
speech	O
;	O
other	O
systems	O
render	O
symbolic	B-General_Concept
linguistic	I-General_Concept
representations	I-General_Concept
like	O
phonetic	O
transcriptions	O
into	O
speech	O
.	O
</s>
<s>
The	O
reverse	O
process	O
is	O
speech	B-Application
recognition	I-Application
.	O
</s>
<s>
Synthesized	B-Application
speech	I-Application
can	O
be	O
created	O
by	O
concatenating	O
pieces	O
of	O
recorded	O
speech	O
that	O
are	O
stored	O
in	O
a	O
database	O
.	O
</s>
<s>
The	O
quality	O
of	O
a	O
speech	B-Application
synthesizer	I-Application
is	O
judged	O
by	O
its	O
similarity	O
to	O
the	O
human	O
voice	O
and	O
by	O
its	O
ability	O
to	O
be	O
understood	O
clearly	O
.	O
</s>
<s>
An	O
intelligible	O
text-to-speech	B-Application
program	O
allows	O
people	O
with	O
visual	O
impairments	O
or	O
reading	O
disabilities	O
to	O
listen	O
to	O
written	O
words	O
on	O
a	O
home	O
computer	O
.	O
</s>
<s>
Many	O
computer	B-General_Concept
operating	I-General_Concept
systems	I-General_Concept
have	O
included	O
speech	B-Application
synthesizers	I-Application
since	O
the	O
early	O
1990s	O
.	O
</s>
<s>
A	O
text-to-speech	B-Application
system	O
(	O
or	O
"	O
engine	O
"	O
)	O
is	O
composed	O
of	O
two	O
parts	O
:	O
a	O
front-end	B-Architecture
and	O
a	O
back-end	B-Architecture
.	O
</s>
<s>
The	O
front-end	B-Architecture
has	O
two	O
major	O
tasks	O
.	O
</s>
<s>
This	O
process	O
is	O
often	O
called	O
text	B-General_Concept
normalization	I-General_Concept
,	O
pre-processing	O
,	O
or	O
tokenization	O
.	O
</s>
<s>
The	O
front-end	B-Architecture
then	O
assigns	O
phonetic	O
transcriptions	O
to	O
each	O
word	O
,	O
and	O
divides	O
and	O
marks	O
the	O
text	O
into	O
prosodic	O
units	O
,	O
like	O
phrases	O
,	O
clauses	O
,	O
and	O
sentences	O
.	O
</s>
<s>
The	O
process	O
of	O
assigning	O
phonetic	O
transcriptions	O
to	O
words	O
is	O
called	O
text-to-phoneme	O
or	O
grapheme-to-phoneme	O
conversion	O
.	O
</s>
<s>
Phonetic	O
transcriptions	O
and	O
prosody	O
information	O
together	O
make	O
up	O
the	O
symbolic	B-General_Concept
linguistic	I-General_Concept
representation	I-General_Concept
that	O
is	O
output	O
by	O
the	O
front-end	B-Architecture
.	O
</s>
<s>
The	O
back-end	B-Architecture
—	O
often	O
referred	O
to	O
as	O
the	O
synthesizer	O
—	O
then	O
converts	O
the	O
symbolic	B-General_Concept
linguistic	I-General_Concept
representation	I-General_Concept
into	O
sound	O
.	O
</s>
<s>
In	O
certain	O
systems	O
,	O
this	O
part	O
includes	O
the	O
computation	O
of	O
the	O
target	O
prosody	O
(	O
pitch	O
contour	O
,	O
phoneme	B-Language
durations	O
)	O
,	O
which	O
is	O
then	O
imposed	O
on	O
the	O
output	O
speech	O
.	O
</s>
<s>
There	O
followed	O
the	O
bellows-operated	O
"	O
acoustic-mechanical	O
speech	O
machine	O
"	O
of	O
Wolfgang	O
von	O
Kempelen	O
of	O
Pressburg	O
,	O
Hungary	O
,	O
described	O
in	O
a	O
1791	O
paper	O
.	O
</s>
<s>
In	O
1837	O
,	O
Charles	O
Wheatstone	O
produced	O
a	O
"	O
speaking	O
machine	O
"	O
based	O
on	O
von	O
Kempelen	O
's	O
design	O
,	O
and	O
in	O
1846	O
,	O
Joseph	O
Faber	O
exhibited	O
the	O
"	O
Euphonia	B-Application
"	O
.	O
</s>
<s>
In	O
the	O
1930s	O
Bell	O
Labs	O
developed	O
the	O
vocoder	B-Application
,	O
which	O
automatically	O
analyzed	O
speech	O
into	O
its	O
fundamental	O
tones	O
and	O
resonances	O
.	O
</s>
<s>
From	O
his	O
work	O
on	O
the	O
vocoder	B-Application
,	O
Homer	O
Dudley	O
developed	O
a	O
keyboard-operated	O
voice-synthesizer	O
called	O
The	B-Application
Voder	I-Application
(	O
Voice	O
Demonstrator	O
)	O
,	O
which	O
he	O
exhibited	O
at	O
the	O
1939	O
New	O
York	O
World	O
's	O
Fair	O
.	O
</s>
<s>
Dr.	O
Franklin	O
S	O
.	O
Cooper	O
and	O
his	O
colleagues	O
at	O
Haskins	O
Laboratories	O
built	O
the	O
Pattern	B-Application
playback	I-Application
in	O
the	O
late	O
1940s	O
and	O
completed	O
it	O
in	O
1950	O
.	O
</s>
<s>
There	O
were	O
several	O
different	O
versions	O
of	O
this	O
hardware	B-Architecture
device	O
;	O
only	O
one	O
currently	O
survives	O
.	O
</s>
<s>
developed	O
the	O
first	O
general	O
English	O
text-to-speech	B-Application
system	O
in	O
1968	O
,	O
at	O
the	O
Electrotechnical	O
Laboratory	O
in	O
Japan	O
.	O
</s>
<s>
In	O
1961	O
,	O
physicist	O
John	O
Larry	O
Kelly	O
,	O
Jr	O
and	O
his	O
colleague	O
Louis	O
Gerstman	O
used	O
an	O
IBM	B-Device
704	I-Device
computer	O
to	O
synthesize	B-Application
speech	I-Application
,	O
an	O
event	O
among	O
the	O
most	O
prominent	O
in	O
the	O
history	O
of	O
Bell	O
Labs	O
.	O
</s>
<s>
Kelly	O
's	O
voice	O
recorder	O
synthesizer	O
(	O
vocoder	B-Application
)	O
recreated	O
the	O
song	O
"	O
Daisy	O
Bell	O
"	O
,	O
with	O
musical	O
accompaniment	O
from	O
Max	O
Mathews	O
.	O
</s>
<s>
Clarke	O
was	O
so	O
impressed	O
by	O
the	O
demonstration	O
that	O
he	O
used	O
it	O
in	O
the	O
climactic	O
scene	O
of	O
his	O
screenplay	O
for	O
his	O
novel	O
2001	O
:	O
A	O
Space	O
Odyssey	O
,	O
where	O
the	O
HAL	B-General_Concept
9000	I-General_Concept
computer	O
sings	O
the	O
same	O
song	O
as	O
astronaut	O
Dave	O
Bowman	O
puts	O
it	O
to	O
sleep	O
.	O
</s>
<s>
Despite	O
the	O
success	O
of	O
purely	O
electronic	O
speech	B-Application
synthesis	I-Application
,	O
research	O
into	O
mechanical	O
speech-synthesizers	O
continues	O
.	O
</s>
<s>
Linear	B-Algorithm
predictive	I-Algorithm
coding	I-Algorithm
(	O
LPC	O
)	O
,	O
a	O
form	O
of	O
speech	B-Application
coding	I-Application
,	O
began	O
development	O
with	O
the	O
work	O
of	O
Fumitada	O
Itakura	O
of	O
Nagoya	O
University	O
and	O
Shuzo	O
Saito	O
of	O
Nippon	O
Telegraph	O
and	O
Telephone	O
(	O
NTT	O
)	O
in	O
1966	O
.	O
</s>
<s>
LPC	O
was	O
later	O
the	O
basis	O
for	O
early	O
speech	B-Application
synthesizer	I-Application
chips	O
,	O
such	O
as	O
the	O
Texas	B-Application
Instruments	I-Application
LPC	I-Application
Speech	I-Application
Chips	I-Application
used	O
in	O
the	O
Speak	B-Application
&	I-Application
Spell	I-Application
toys	O
from	O
1978	O
.	O
</s>
<s>
In	O
1975	O
,	O
Fumitada	O
Itakura	O
developed	O
the	O
line	B-Algorithm
spectral	I-Algorithm
pairs	I-Algorithm
(	O
LSP	O
)	O
method	O
for	O
high-compression	O
speech	B-Application
coding	I-Application
,	O
while	O
at	O
NTT	O
.	O
</s>
<s>
In	O
1980	O
,	O
his	O
team	O
developed	O
an	O
LSP-based	O
speech	B-Application
synthesizer	I-Application
chip	O
.	O
</s>
<s>
LSP	O
is	O
an	O
important	O
technology	O
for	O
speech	B-Application
synthesis	I-Application
and	O
coding	O
,	O
and	O
in	O
the	O
1990s	O
was	O
adopted	O
by	O
almost	O
all	O
international	O
speech	B-Application
coding	I-Application
standards	O
as	O
an	O
essential	O
component	O
,	O
contributing	O
to	O
the	O
enhancement	O
of	O
digital	O
speech	O
communication	O
over	O
mobile	O
channels	O
and	O
the	O
internet	O
.	O
</s>
<s>
In	O
1975	O
,	O
MUSA	B-Language
was	O
released	O
,	O
and	O
was	O
one	O
of	O
the	O
first	O
Speech	B-Application
Synthesis	I-Application
systems	O
.	O
</s>
<s>
It	O
consisted	O
of	O
a	O
stand-alone	O
computer	B-Architecture
hardware	I-Architecture
and	O
a	O
specialized	O
software	O
that	O
enabled	O
it	O
to	O
read	O
Italian	O
.	O
</s>
<s>
Dominant	O
systems	O
in	O
the	O
1980s	O
and	O
1990s	O
were	O
the	O
DECtalk	B-Application
system	O
,	O
based	O
largely	O
on	O
the	O
work	O
of	O
Dennis	O
Klatt	O
at	O
MIT	O
,	O
and	O
the	O
Bell	O
Labs	O
system	O
;	O
the	O
latter	O
was	O
one	O
of	O
the	O
first	O
multilingual	O
language-independent	O
systems	O
,	O
making	O
extensive	O
use	O
of	O
natural	B-Language
language	I-Language
processing	I-Language
methods	O
.	O
</s>
<s>
Handheld	B-Application
electronics	I-Application
featuring	O
speech	B-Application
synthesis	I-Application
began	O
emerging	O
in	O
the	O
1970s	O
.	O
</s>
<s>
Other	O
devices	O
had	O
primarily	O
educational	O
purposes	O
,	O
such	O
as	O
the	O
Speak	B-Application
&	I-Application
Spell	I-Application
toy	I-Application
produced	O
by	O
Texas	O
Instruments	O
in	O
1978	O
.	O
</s>
<s>
The	O
first	O
video	O
game	O
to	O
feature	O
speech	B-Application
synthesis	I-Application
was	O
the	O
1980	O
shoot	O
'	O
em	O
up	O
arcade	O
game	O
,	O
Stratovox	B-Application
(	O
known	O
in	O
Japan	O
as	O
Speak	O
&	O
Rescue	O
)	O
,	O
from	O
Sun	O
Electronics	O
.	O
</s>
<s>
The	O
first	O
personal	O
computer	O
game	O
with	O
speech	B-Application
synthesis	I-Application
was	O
Manbiki	O
Shoujo	O
(	O
Shoplifting	O
Girl	O
)	O
,	O
released	O
in	O
1980	O
for	O
the	O
PET	B-Device
2001	I-Device
,	O
for	O
which	O
the	O
game	O
's	O
developer	O
,	O
Hiroshi	O
Suzuki	O
,	O
developed	O
a	O
"	O
zero	O
cross	O
"	O
programming	O
technique	O
to	O
produce	O
a	O
synthesized	B-Application
speech	I-Application
waveform	O
.	O
</s>
<s>
Another	O
early	O
example	O
,	O
the	O
arcade	O
version	O
of	O
Berzerk	B-Application
,	O
also	O
dates	O
from	O
1980	O
.	O
</s>
<s>
The	O
Milton	O
Bradley	O
Company	O
produced	O
the	O
first	O
multi-player	O
electronic	O
game	O
using	O
voice	B-Application
synthesis	I-Application
,	O
Milton	O
,	O
in	O
the	O
same	O
year	O
.	O
</s>
<s>
The	O
quality	O
of	O
synthesized	B-Application
speech	I-Application
has	O
steadily	O
improved	O
,	O
but	O
output	O
from	O
contemporary	O
speech	B-Application
synthesis	I-Application
systems	O
remains	O
clearly	O
distinguishable	O
from	O
actual	O
human	O
speech	O
.	O
</s>
<s>
Kurzweil	O
predicted	O
in	O
2005	O
that	O
as	O
the	O
cost-performance	O
ratio	O
caused	O
speech	B-Application
synthesizers	I-Application
to	O
become	O
cheaper	O
and	O
more	O
accessible	O
,	O
more	O
people	O
would	O
benefit	O
from	O
the	O
use	O
of	O
text-to-speech	B-Application
programs	O
.	O
</s>
<s>
The	O
most	O
important	O
qualities	O
of	O
a	O
speech	B-Application
synthesis	I-Application
system	O
are	O
naturalness	O
and	O
intelligibility	O
.	O
</s>
<s>
The	O
ideal	O
speech	B-Application
synthesizer	I-Application
is	O
both	O
natural	O
and	O
intelligible	O
.	O
</s>
<s>
Speech	B-Application
synthesis	I-Application
systems	O
usually	O
try	O
to	O
maximize	O
both	O
characteristics	O
.	O
</s>
<s>
Generally	O
,	O
concatenative	O
synthesis	O
produces	O
the	O
most	O
natural-sounding	O
synthesized	B-Application
speech	I-Application
.	O
</s>
<s>
During	O
database	O
creation	O
,	O
each	O
recorded	O
utterance	O
is	O
segmented	O
into	O
some	O
or	O
all	O
of	O
the	O
following	O
:	O
individual	O
phones	O
,	O
diphones	O
,	O
half-phones	O
,	O
syllables	B-General_Concept
,	O
morphemes	O
,	O
words	O
,	O
phrases	O
,	O
and	O
sentences	O
.	O
</s>
<s>
Typically	O
,	O
the	O
division	O
into	O
segments	O
is	O
done	O
using	O
a	O
specially	O
modified	O
speech	B-Application
recognizer	I-Application
set	O
to	O
a	O
"	O
forced	O
alignment	O
"	O
mode	O
with	O
some	O
manual	O
correction	O
afterward	O
,	O
using	O
visual	O
representations	O
such	O
as	O
the	O
waveform	O
and	O
spectrogram	O
.	O
</s>
<s>
An	O
index	B-Data_Structure
of	O
the	O
units	O
in	O
the	O
speech	O
database	O
is	O
then	O
created	O
based	O
on	O
the	O
segmentation	O
and	O
acoustic	O
parameters	O
like	O
the	O
fundamental	O
frequency	O
(	O
pitch	O
)	O
,	O
duration	O
,	O
position	O
in	O
the	O
syllable	B-General_Concept
,	O
and	O
neighboring	O
phones	O
.	O
</s>
<s>
At	O
run	B-Library
time	I-Library
,	O
the	O
desired	O
target	O
utterance	O
is	O
created	O
by	O
determining	O
the	O
best	O
chain	O
of	O
candidate	O
units	O
from	O
the	O
database	O
(	O
unit	O
selection	O
)	O
.	O
</s>
<s>
This	O
process	O
is	O
typically	O
achieved	O
using	O
a	O
specially	O
weighted	O
decision	B-Algorithm
tree	I-Algorithm
.	O
</s>
<s>
Unit	O
selection	O
provides	O
the	O
greatest	O
naturalness	O
,	O
because	O
it	O
applies	O
only	O
a	O
small	O
amount	O
of	O
digital	B-General_Concept
signal	I-General_Concept
processing	I-General_Concept
(	O
DSP	O
)	O
to	O
the	O
recorded	O
speech	O
.	O
</s>
<s>
Recently	O
,	O
researchers	O
have	O
proposed	O
various	O
automated	O
methods	O
to	O
detect	O
unnatural	O
segments	O
in	O
unit-selection	O
speech	B-Application
synthesis	I-Application
systems	O
.	O
</s>
<s>
At	B-Library
runtime	I-Library
,	O
the	O
target	O
prosody	O
of	O
a	O
sentence	O
is	O
superimposed	O
on	O
these	O
minimal	O
units	O
by	O
means	O
of	O
digital	B-General_Concept
signal	I-General_Concept
processing	I-General_Concept
techniques	O
such	O
as	O
linear	B-Algorithm
predictive	I-Algorithm
coding	I-Algorithm
,	O
PSOLA	B-Application
or	O
MBROLA	B-Application
.	O
</s>
<s>
or	O
more	O
recent	O
techniques	O
such	O
as	O
pitch	O
modification	O
in	O
the	O
source	O
domain	O
using	O
discrete	B-General_Concept
cosine	I-General_Concept
transform	I-General_Concept
.	O
</s>
<s>
As	O
such	O
,	O
its	O
use	O
in	O
commercial	O
applications	B-Application
is	O
declining	O
,	O
although	O
it	O
continues	O
to	O
be	O
used	O
in	O
research	O
because	O
there	O
are	O
a	O
number	O
of	O
freely	O
available	O
software	O
implementations	O
.	O
</s>
<s>
It	O
is	O
used	O
in	O
applications	B-Application
where	O
the	O
variety	O
of	O
texts	O
the	O
system	O
will	O
output	O
is	O
limited	O
to	O
a	O
particular	O
domain	O
,	O
like	O
transit	O
schedule	O
announcements	O
or	O
weather	O
reports	O
.	O
</s>
<s>
Formant	O
synthesis	O
does	O
not	O
use	O
human	O
speech	O
samples	O
at	B-Library
runtime	I-Library
.	O
</s>
<s>
Instead	O
,	O
the	O
synthesized	B-Application
speech	I-Application
output	O
is	O
created	O
using	O
additive	O
synthesis	O
and	O
an	O
acoustic	O
model	O
(	O
physical	O
modelling	O
synthesis	O
)	O
.	O
</s>
<s>
However	O
,	O
maximum	O
naturalness	O
is	O
not	O
always	O
the	O
goal	O
of	O
a	O
speech	B-Application
synthesis	I-Application
system	O
,	O
and	O
formant	O
synthesis	O
systems	O
have	O
advantages	O
over	O
concatenative	O
systems	O
.	O
</s>
<s>
High-speed	O
synthesized	B-Application
speech	I-Application
is	O
used	O
by	O
the	O
visually	O
impaired	O
to	O
quickly	O
navigate	O
computers	O
using	O
a	O
screen	O
reader	O
.	O
</s>
<s>
They	O
can	O
therefore	O
be	O
used	O
in	O
embedded	B-Architecture
systems	I-Architecture
,	O
where	O
memory	B-General_Concept
and	O
microprocessor	B-Architecture
power	O
are	O
especially	O
limited	O
.	O
</s>
<s>
Examples	O
of	O
non-real-time	O
but	O
highly	O
accurate	O
intonation	O
control	O
in	O
formant	O
synthesis	O
include	O
the	O
work	O
done	O
in	O
the	O
late	O
1970s	O
for	O
the	O
Texas	O
Instruments	O
toy	O
Speak	B-Application
&	I-Application
Spell	I-Application
,	O
and	O
in	O
the	O
early	O
1980s	O
Sega	O
arcade	O
machines	O
and	O
in	O
many	O
Atari	O
,	O
Inc	O
.	O
arcade	O
games	O
using	O
the	O
TMS5220	B-Application
LPC	I-Application
Chips	I-Application
.	O
</s>
<s>
Creating	O
proper	O
intonation	O
for	O
these	O
projects	O
was	O
painstaking	O
,	O
and	O
the	O
results	O
have	O
yet	O
to	O
be	O
matched	O
by	O
real-time	O
text-to-speech	B-Application
interfaces	O
.	O
</s>
<s>
Articulatory	B-Application
synthesis	I-Application
refers	O
to	O
computational	O
techniques	O
for	O
synthesizing	O
speech	O
based	O
on	O
models	O
of	O
the	O
human	O
vocal	O
tract	O
and	O
the	O
articulation	O
processes	O
occurring	O
there	O
.	O
</s>
<s>
Until	O
recently	O
,	O
articulatory	B-Application
synthesis	I-Application
models	O
have	O
not	O
been	O
incorporated	O
into	O
commercial	O
speech	B-Application
synthesis	I-Application
systems	O
.	O
</s>
<s>
Following	O
the	O
demise	O
of	O
the	O
various	O
incarnations	O
of	O
NeXT	O
(	O
started	O
by	O
Steve	O
Jobs	O
in	O
the	O
late	O
1980s	O
and	O
merged	O
with	O
Apple	O
Computer	O
in	O
1997	O
)	O
,	O
the	O
Trillium	O
software	O
was	O
published	O
under	O
the	O
GNU	O
General	O
Public	O
License	O
,	O
with	O
work	O
continuing	O
as	O
gnuspeech	B-Operating_System
.	O
</s>
<s>
The	O
system	O
,	O
first	O
marketed	O
in	O
1994	O
,	O
provides	O
full	O
articulatory-based	O
text-to-speech	B-Application
conversion	O
using	O
a	O
waveguide	O
or	O
transmission-line	O
analog	O
of	O
the	O
human	O
oral	O
and	O
nasal	O
tracts	O
controlled	O
by	O
Carré	O
's	O
"	O
distinctive	O
region	O
model	O
"	O
.	O
</s>
<s>
More	O
recent	O
synthesizers	O
,	O
developed	O
by	O
Jorge	O
C	O
.	O
Lucero	O
and	O
colleagues	O
,	O
incorporate	O
models	O
of	O
vocal	O
fold	O
biomechanics	O
,	O
glottal	O
aerodynamics	O
and	O
acoustic	O
wave	O
propagation	O
in	O
the	O
bronchi	O
,	O
trachea	O
,	O
nasal	O
and	O
oral	O
cavities	O
,	O
and	O
thus	O
constitute	O
full	O
systems	O
of	O
physics-based	O
speech	B-Application
simulation	I-Application
.	O
</s>
<s>
Sinewave	B-Application
synthesis	I-Application
is	O
a	O
technique	O
for	O
synthesizing	O
speech	O
by	O
replacing	O
the	O
formants	O
(	O
main	O
bands	O
of	O
energy	O
)	O
with	O
pure	O
tone	O
whistles	O
.	O
</s>
<s>
artificial	O
speech	O
from	O
text	O
(	O
text-to-speech	B-Application
)	O
or	O
spectrum	O
(	O
vocoder	B-Application
)	O
.	O
</s>
<s>
a	O
text-to-speech	B-Application
system	O
,	O
the	O
associated	O
labels	O
and/or	O
input	O
text	O
.	O
</s>
<s>
The	O
DNN-based	O
speech	B-Application
synthesizers	I-Application
are	O
approaching	O
the	O
naturalness	O
of	O
the	O
human	O
voice	O
.	O
</s>
<s>
For	O
tonal	O
languages	O
,	O
such	O
as	O
Chinese	O
or	O
Taiwanese	O
language	O
,	O
there	O
are	O
different	O
levels	O
of	O
tone	O
sandhi	O
required	O
and	O
sometimes	O
the	O
output	O
of	O
speech	B-Application
synthesizer	I-Application
may	O
result	O
in	O
the	O
mistakes	O
of	O
tone	O
sandhi	O
.	O
</s>
<s>
Most	O
text-to-speech	B-Application
(	O
TTS	O
)	O
systems	O
do	O
not	O
generate	O
semantic	O
representations	O
of	O
their	O
input	O
texts	O
,	O
as	O
processes	O
for	O
doing	O
so	O
are	O
unreliable	O
,	O
poorly	O
understood	O
,	O
and	O
computationally	O
ineffective	O
.	O
</s>
<s>
As	O
a	O
result	O
,	O
various	O
heuristic	B-Algorithm
techniques	O
are	O
used	O
to	O
guess	O
the	O
proper	O
way	O
to	O
disambiguate	O
homographs	O
,	O
like	O
examining	O
neighboring	O
words	O
and	O
using	O
statistics	O
about	O
frequency	O
of	O
occurrence	O
.	O
</s>
<s>
TTS	O
systems	O
with	O
intelligent	O
front	B-Architecture
ends	I-Architecture
can	O
make	O
educated	O
guesses	O
about	O
ambiguous	O
abbreviations	O
,	O
while	O
others	O
provide	O
the	O
same	O
result	O
in	O
all	O
cases	O
,	O
resulting	O
in	O
nonsensical	O
(	O
and	O
sometimes	O
comical	O
)	O
outputs	O
,	O
such	O
as	O
"	O
Ulysses	O
S	O
.	O
Grant	O
"	O
being	O
rendered	O
as	O
"	O
Ulysses	O
South	O
Grant	O
"	O
.	O
</s>
<s>
Speech	B-Application
synthesis	I-Application
systems	O
use	O
two	O
basic	O
approaches	O
to	O
determine	O
the	O
pronunciation	O
of	O
a	O
word	O
based	O
on	O
its	O
spelling	O
,	O
a	O
process	O
which	O
is	O
often	O
called	O
text-to-phoneme	O
or	O
grapheme-to-phoneme	O
conversion	O
(	O
phoneme	B-Language
is	O
the	O
term	O
used	O
by	O
linguists	O
to	O
describe	O
distinctive	O
sounds	O
in	O
a	O
language	O
)	O
.	O
</s>
<s>
The	O
simplest	O
approach	O
to	O
text-to-phoneme	O
conversion	O
is	O
the	O
dictionary-based	O
approach	O
,	O
where	O
a	O
large	O
dictionary	O
containing	O
all	O
the	O
words	O
of	O
a	O
language	O
and	O
their	O
correct	O
pronunciations	O
is	O
stored	O
by	O
the	O
program	O
.	O
</s>
<s>
As	O
dictionary	O
size	O
grows	O
,	O
so	O
too	O
does	O
the	O
memory	B-General_Concept
space	O
requirements	O
of	O
the	O
synthesis	O
system	O
.	O
</s>
<s>
As	O
a	O
result	O
,	O
nearly	O
all	O
speech	B-Application
synthesis	I-Application
systems	O
use	O
a	O
combination	O
of	O
these	O
approaches	O
.	O
</s>
<s>
Speech	B-Application
synthesis	I-Application
systems	O
for	O
such	O
languages	O
often	O
use	O
the	O
rule-based	O
method	O
extensively	O
,	O
resorting	O
to	O
dictionaries	O
only	O
for	O
those	O
few	O
words	O
,	O
like	O
foreign	O
names	O
and	O
loanwords	O
,	O
whose	O
pronunciations	O
are	O
not	O
obvious	O
from	O
their	O
spellings	O
.	O
</s>
<s>
On	O
the	O
other	O
hand	O
,	O
speech	B-Application
synthesis	I-Application
systems	O
for	O
languages	O
like	O
English	O
,	O
which	O
have	O
extremely	O
irregular	O
spelling	O
systems	O
,	O
are	O
more	O
likely	O
to	O
rely	O
on	O
dictionaries	O
,	O
and	O
to	O
use	O
rule-based	O
methods	O
only	O
for	O
unusual	O
words	O
,	O
or	O
words	O
that	O
are	O
n't	O
in	O
their	O
dictionaries	O
.	O
</s>
<s>
The	O
consistent	O
evaluation	O
of	O
speech	B-Application
synthesis	I-Application
systems	O
may	O
be	O
difficult	O
because	O
of	O
a	O
lack	O
of	O
universally	O
agreed	O
objective	O
evaluation	O
criteria	O
.	O
</s>
<s>
The	O
quality	O
of	O
speech	B-Application
synthesis	I-Application
systems	O
also	O
depends	O
on	O
the	O
quality	O
of	O
the	O
production	O
technique	O
(	O
which	O
may	O
involve	O
analogue	O
or	O
digital	O
recording	O
)	O
and	O
on	O
the	O
facilities	O
used	O
to	O
replay	O
the	O
speech	O
.	O
</s>
<s>
Evaluating	O
speech	B-Application
synthesis	I-Application
systems	O
has	O
therefore	O
often	O
been	O
compromised	O
by	O
differences	O
between	O
production	O
techniques	O
and	O
replay	O
facilities	O
.	O
</s>
<s>
Since	O
2005	O
,	O
however	O
,	O
some	O
researchers	O
have	O
started	O
to	O
evaluate	O
speech	B-Application
synthesis	I-Application
systems	O
using	O
a	O
common	O
speech	O
dataset	O
.	O
</s>
<s>
It	O
was	O
suggested	O
that	O
identification	O
of	O
the	O
vocal	O
features	O
that	O
signal	O
emotional	O
content	O
may	O
be	O
used	O
to	O
help	O
make	O
synthesized	B-Application
speech	I-Application
sound	O
more	O
natural	O
.	O
</s>
<s>
One	O
of	O
the	O
techniques	O
for	O
pitch	O
modification	O
uses	O
discrete	B-General_Concept
cosine	I-General_Concept
transform	I-General_Concept
in	O
the	O
source	O
domain	O
(	O
linear	O
prediction	O
residual	O
)	O
.	O
</s>
<s>
Such	O
pitch	O
synchronous	O
pitch	O
modification	O
techniques	O
need	O
a	O
priori	O
pitch	O
marking	O
of	O
the	O
synthesis	O
speech	O
database	O
using	O
techniques	O
such	O
as	O
epoch	O
extraction	O
using	O
dynamic	O
plosion	O
index	B-Data_Structure
applied	O
on	O
the	O
integrated	O
linear	O
prediction	O
residual	O
of	O
the	O
voiced	O
regions	O
of	O
speech	O
.	O
</s>
<s>
Popular	O
systems	O
offering	O
speech	B-Application
synthesis	I-Application
as	O
a	O
built-in	O
capability	O
.	O
</s>
<s>
In	O
the	O
early	O
1980s	O
,	O
TI	O
was	O
known	O
as	O
a	O
pioneer	O
in	O
speech	B-Application
synthesis	I-Application
,	O
and	O
a	O
highly	O
popular	O
plug-in	B-Application
speech	B-Application
synthesizer	I-Application
module	O
was	O
available	O
for	O
the	O
TI-99/4	B-Device
and	O
4A	O
.	O
</s>
<s>
Speech	B-Application
synthesizers	I-Application
were	O
offered	O
free	O
with	O
the	O
purchase	O
of	O
a	O
number	O
of	O
cartridges	O
and	O
were	O
used	O
by	O
many	O
TI-written	O
video	O
games	O
(	O
games	O
offered	O
with	O
speech	O
during	O
this	O
promotion	O
included	O
Alpiner	B-Application
and	O
Parsec	B-Application
)	O
.	O
</s>
<s>
The	O
synthesizer	O
uses	O
a	O
variant	O
of	O
linear	B-Algorithm
predictive	I-Algorithm
coding	I-Algorithm
and	O
has	O
a	O
small	O
in-built	O
vocabulary	O
.	O
</s>
<s>
However	O
,	O
the	O
success	O
of	O
software	O
text-to-speech	B-Application
in	O
the	O
Terminal	O
Emulator	O
II	O
cartridge	O
canceled	O
that	O
plan	O
.	O
</s>
<s>
The	O
Mattel	B-Operating_System
Intellivision	I-Operating_System
game	O
console	O
offered	O
the	O
Intellivoice	O
Voice	B-Application
Synthesis	I-Application
module	O
in	O
1982	O
.	O
</s>
<s>
It	O
included	O
the	O
SP0256	B-Application
Narrator	I-Application
speech	B-Application
synthesizer	I-Application
chip	O
on	O
a	O
removable	O
cartridge	O
.	O
</s>
<s>
The	O
Narrator	O
had	O
2kB	O
of	O
Read-Only	O
Memory	B-General_Concept
(	O
ROM	O
)	O
,	O
and	O
this	O
was	O
utilized	O
to	O
store	O
a	O
database	O
of	O
generic	O
words	O
that	O
could	O
be	O
combined	O
to	O
make	O
phrases	O
in	O
Intellivision	B-Operating_System
games	O
.	O
</s>
<s>
Since	O
the	O
Orator	O
chip	O
could	O
also	O
accept	O
speech	O
data	O
from	O
external	O
memory	B-General_Concept
,	O
any	O
additional	O
words	O
or	O
phrases	O
needed	O
could	O
be	O
stored	O
inside	O
the	O
cartridge	O
itself	O
.	O
</s>
<s>
Also	O
released	O
in	O
1982	O
,	O
Software	B-Application
Automatic	I-Application
Mouth	I-Application
was	O
the	O
first	O
commercial	O
all-software	O
voice	B-Application
synthesis	I-Application
program	O
.	O
</s>
<s>
It	O
was	O
later	O
used	O
as	O
the	O
basis	O
for	O
Macintalk	B-Operating_System
.	O
</s>
<s>
The	O
program	O
was	O
available	O
for	O
non-Macintosh	O
Apple	O
computers	O
(	O
including	O
the	O
Apple	O
II	O
,	O
and	O
the	O
Lisa	O
)	O
,	O
various	O
Atari	O
models	O
and	O
the	O
Commodore	O
64	O
.	O
</s>
<s>
The	O
Apple	O
version	O
preferred	O
additional	O
hardware	B-Architecture
that	O
contained	O
DACs	O
,	O
although	O
it	O
could	O
instead	O
use	O
the	O
computer	O
's	O
one-bit	O
audio	O
output	O
(	O
with	O
the	O
addition	O
of	O
much	O
distortion	O
)	O
if	O
the	O
card	O
was	O
not	O
present	O
.	O
</s>
<s>
Arguably	O
,	O
the	O
first	O
speech	O
system	O
integrated	O
into	O
an	O
operating	B-General_Concept
system	I-General_Concept
was	O
the	O
1400XL/1450XL	O
personal	O
computers	O
designed	O
by	O
Atari	O
,	O
Inc	O
.	O
using	O
the	O
Votrax	B-Application
SC01	O
chip	O
in	O
1983	O
.	O
</s>
<s>
The	O
1400XL/1450XL	O
computers	O
used	O
a	O
Finite	O
State	O
Machine	O
to	O
enable	O
World	O
English	O
Spelling	O
text-to-speech	B-Application
synthesis	O
.	O
</s>
<s>
The	O
Atari	B-Device
ST	I-Device
computers	O
were	O
sold	O
with	O
"	O
stspeech.tos	O
"	O
on	O
floppy	O
disk	O
.	O
</s>
<s>
The	O
first	O
speech	O
system	O
integrated	O
into	O
an	O
operating	B-General_Concept
system	I-General_Concept
that	O
shipped	O
in	O
quantity	O
was	O
Apple	O
Computer	O
's	O
MacInTalk	B-Operating_System
.	O
</s>
<s>
The	O
software	O
was	O
licensed	O
from	O
third-party	O
developers	O
Joseph	O
Katz	O
and	O
Mark	O
Barton	O
(	O
later	O
,	O
SoftVoice	O
,	O
Inc	O
.	O
)	O
and	O
was	O
featured	O
during	O
the	O
1984	O
introduction	O
of	O
the	O
Macintosh	B-Device
computer	O
.	O
</s>
<s>
This	O
January	O
demo	O
required	O
512	O
kilobytes	O
of	O
RAM	O
memory	B-General_Concept
.	O
</s>
<s>
So	O
,	O
the	O
demo	O
was	O
accomplished	O
with	O
a	O
prototype	O
512k	O
Mac	O
,	O
although	O
those	O
in	O
attendance	O
were	O
not	O
told	O
of	O
this	O
and	O
the	O
synthesis	O
demo	O
created	O
considerable	O
excitement	O
for	O
the	O
Macintosh	B-Device
.	O
</s>
<s>
In	O
the	O
early	O
1990s	O
Apple	O
expanded	O
its	O
capabilities	O
offering	O
system	O
wide	O
text-to-speech	B-Application
support	O
.	O
</s>
<s>
Apple	O
also	O
introduced	O
speech	B-Application
recognition	I-Application
into	O
its	O
systems	O
which	O
provided	O
a	O
fluid	O
command	O
set	O
.	O
</s>
<s>
Starting	O
as	O
a	O
curiosity	O
,	O
the	O
speech	O
system	O
of	O
Apple	O
Macintosh	B-Device
has	O
evolved	O
into	O
a	O
fully	O
supported	O
program	O
,	O
PlainTalk	B-Operating_System
,	O
for	O
people	O
with	O
vision	O
problems	O
.	O
</s>
<s>
VoiceOver	B-Operating_System
was	O
for	O
the	O
first	O
time	O
featured	O
in	O
2005	O
in	O
Mac	B-Application
OS	I-Application
X	I-Application
Tiger	I-Application
(	O
10.4	B-Application
)	O
.	O
</s>
<s>
During	O
10.4	B-Application
(	O
Tiger	O
)	O
and	O
first	O
releases	O
of	O
10.5	O
(	O
Leopard	B-Application
)	O
there	O
was	O
only	O
one	O
standard	O
voice	O
shipping	O
with	O
Mac	B-Device
OS	I-Device
X	I-Device
.	O
</s>
<s>
Starting	O
with	O
10.6	B-Device
(	O
Snow	B-Device
Leopard	I-Device
)	O
,	O
the	O
user	O
can	O
choose	O
out	O
of	O
a	O
wide	O
range	O
list	O
of	O
multiple	O
voices	O
.	O
</s>
<s>
VoiceOver	B-Operating_System
voices	O
feature	O
the	O
taking	O
of	O
realistic-sounding	O
breaths	O
between	O
sentences	O
,	O
as	O
well	O
as	O
improved	O
clarity	O
at	O
high	O
read	O
rates	O
over	O
PlainTalk	B-Operating_System
.	O
</s>
<s>
Mac	B-Device
OS	I-Device
X	I-Device
also	O
includes	O
say	O
,	O
a	O
command-line	B-Application
based	I-Application
application	O
that	O
converts	O
text	O
to	O
audible	O
speech	O
.	O
</s>
<s>
The	O
AppleScript	B-Application
Standard	O
Additions	O
includes	O
a	O
say	O
verb	O
that	O
allows	O
a	O
script	O
to	O
use	O
any	O
of	O
the	O
installed	O
voices	O
and	O
to	O
control	O
the	O
pitch	O
,	O
speaking	O
rate	O
and	O
modulation	O
of	O
the	O
spoken	O
text	O
.	O
</s>
<s>
Used	O
in	O
Alexa	B-Application
and	O
as	O
Software	B-Architecture
as	I-Architecture
a	I-Architecture
Service	I-Architecture
in	O
AWS	O
(	O
from	O
2017	O
)	O
.	O
</s>
<s>
The	O
second	O
operating	B-General_Concept
system	I-General_Concept
to	O
feature	O
advanced	O
speech	B-Application
synthesis	I-Application
capabilities	O
was	O
AmigaOS	B-Application
,	O
introduced	O
in	O
1985	O
.	O
</s>
<s>
The	O
voice	B-Application
synthesis	I-Application
was	O
licensed	O
by	O
Commodore	O
International	O
from	O
SoftVoice	O
,	O
Inc.	O
,	O
who	O
also	O
developed	O
the	O
original	O
MacinTalk	B-Operating_System
text-to-speech	B-Application
system	O
.	O
</s>
<s>
It	O
featured	O
a	O
complete	O
system	O
of	O
voice	O
emulation	O
for	O
American	O
English	O
,	O
with	O
both	O
male	O
and	O
female	O
voices	O
and	O
"	O
stress	O
"	O
indicator	O
markers	O
,	O
made	O
possible	O
through	O
the	O
Amiga	B-Device
's	O
audio	O
chipset	B-Device
.	O
</s>
<s>
The	O
synthesis	O
system	O
was	O
divided	O
into	O
a	O
translator	O
library	O
which	O
converted	O
unrestricted	O
English	O
text	O
into	O
a	O
standard	O
set	O
of	O
phonetic	O
codes	O
and	O
a	O
narrator	O
device	O
which	O
implemented	O
a	O
formant	O
model	O
of	O
speech	O
generation	O
..	O
AmigaOS	B-Application
also	O
featured	O
a	O
high-level	O
"	O
Speak	O
Handler	O
"	O
,	O
which	O
allowed	O
command-line	B-Application
users	O
to	O
redirect	O
text	O
output	O
to	O
speech	O
.	O
</s>
<s>
Speech	B-Application
synthesis	I-Application
was	O
occasionally	O
used	O
in	O
third-party	O
programs	O
,	O
particularly	O
word	O
processors	O
and	O
educational	O
software	O
.	O
</s>
<s>
The	O
synthesis	O
software	O
remained	O
largely	O
unchanged	O
from	O
the	O
first	O
AmigaOS	B-Application
release	O
and	O
Commodore	O
eventually	O
removed	O
speech	B-Application
synthesis	I-Application
support	O
from	O
AmigaOS	B-Application
2.1	O
onward	O
.	O
</s>
<s>
Despite	O
the	O
American	O
English	O
phoneme	B-Language
limitation	O
,	O
an	O
unofficial	O
version	O
with	O
multilingual	O
speech	B-Application
synthesis	I-Application
was	O
developed	O
.	O
</s>
<s>
Modern	O
Windows	B-Application
desktop	B-General_Concept
systems	I-General_Concept
can	O
use	O
SAPI	O
4	O
and	O
SAPI	O
5	O
components	O
to	O
support	O
speech	B-Application
synthesis	I-Application
and	O
speech	B-Application
recognition	I-Application
.	O
</s>
<s>
SAPI	O
4.0	O
was	O
available	O
as	O
an	O
optional	O
add-on	O
for	O
Windows	B-Application
95	I-Application
and	O
Windows	B-Device
98	I-Device
.	O
</s>
<s>
Windows	B-Application
2000	I-Application
added	O
Narrator	O
,	O
a	O
text-to-speech	B-Application
utility	O
for	O
people	O
who	O
have	O
visual	O
impairment	O
.	O
</s>
<s>
Third-party	O
programs	O
such	O
as	O
JAWS	O
for	O
Windows	B-Application
,	O
Window-Eyes	O
,	O
Non-visual	O
Desktop	O
Access	O
,	O
Supernova	O
and	O
System	O
Access	O
can	O
perform	O
various	O
text-to-speech	B-Application
tasks	O
such	O
as	O
reading	O
text	O
aloud	O
from	O
a	O
specified	O
website	O
,	O
email	O
account	O
,	O
text	O
document	O
,	O
the	O
Windows	B-Application
clipboard	O
,	O
the	O
user	O
's	O
keyboard	O
typing	O
,	O
etc	O
.	O
</s>
<s>
Not	O
all	O
programs	O
can	O
use	O
speech	B-Application
synthesis	I-Application
directly	O
.	O
</s>
<s>
Some	O
programs	O
can	O
use	O
plug-ins	B-Application
,	O
extensions	O
or	O
add-ons	O
to	O
read	O
text	O
aloud	O
.	O
</s>
<s>
Microsoft	B-Application
Speech	I-Application
Server	I-Application
is	O
a	O
server-based	O
package	O
for	O
voice	B-Application
synthesis	I-Application
and	O
recognition	O
.	O
</s>
<s>
It	O
is	O
designed	O
for	O
network	O
use	O
with	O
web	B-Application
applications	I-Application
and	O
call	O
centers	O
.	O
</s>
<s>
From	O
1971	O
to	O
1996	O
,	O
Votrax	B-Application
produced	O
a	O
number	O
of	O
commercial	O
speech	B-Application
synthesizer	I-Application
components	O
.	O
</s>
<s>
A	O
Votrax	B-Application
synthesizer	O
was	O
included	O
in	O
the	O
first	O
generation	O
Kurzweil	O
Reading	O
Machine	O
for	O
the	O
Blind	O
.	O
</s>
<s>
Text-to-speech	B-Application
(	O
TTS	O
)	O
refers	O
to	O
the	O
ability	O
of	O
computers	O
to	O
read	O
text	O
aloud	O
.	O
</s>
<s>
Version	O
1.6	O
of	O
Android	B-Application
added	O
support	O
for	O
speech	B-Application
synthesis	I-Application
(	O
TTS	O
)	O
.	O
</s>
<s>
Currently	O
,	O
there	O
are	O
a	O
number	O
of	O
applications	B-Application
,	O
plugins	B-Application
and	O
gadgets	O
that	O
can	O
read	O
messages	O
directly	O
from	O
an	O
e-mail	B-Protocol
client	I-Protocol
and	O
web	O
pages	O
from	O
a	O
web	B-Application
browser	I-Application
or	O
Google	B-Protocol
Toolbar	I-Protocol
.	O
</s>
<s>
On	O
one	O
hand	O
,	O
online	O
RSS-narrators	O
simplify	O
information	O
delivery	O
by	O
allowing	O
users	O
to	O
listen	O
to	O
their	O
favourite	O
news	O
sources	O
and	O
to	O
convert	O
them	O
to	O
podcasts	B-Operating_System
.	O
</s>
<s>
Users	O
can	O
download	O
generated	O
audio	O
files	O
to	O
portable	B-Application
devices	I-Application
,	O
e.g.	O
</s>
<s>
with	O
a	O
help	O
of	O
podcast	B-Operating_System
receiver	O
,	O
and	O
listen	O
to	O
them	O
while	O
walking	O
,	O
jogging	O
or	O
commuting	O
to	O
work	O
.	O
</s>
<s>
A	O
growing	O
field	O
in	O
Internet	O
based	O
TTS	O
is	O
web-based	B-Application
assistive	O
technology	O
,	O
e.g.	O
</s>
<s>
It	O
can	O
deliver	O
TTS	O
functionality	O
to	O
anyone	O
(	O
for	O
reasons	O
of	O
accessibility	O
,	O
convenience	O
,	O
entertainment	O
or	O
information	O
)	O
with	O
access	O
to	O
a	O
web	B-Application
browser	I-Application
.	O
</s>
<s>
The	O
non-profit	O
project	O
Pediaphon	O
was	O
created	O
in	O
2006	O
to	O
provide	O
a	O
similar	O
web-based	B-Application
TTS	O
interface	O
to	O
the	O
Wikipedia	O
.	O
</s>
<s>
Other	O
work	O
is	O
being	O
done	O
in	O
the	O
context	O
of	O
the	O
W3C	O
through	O
the	O
W3C	O
Audio	O
Incubator	O
Group	O
with	O
the	O
involvement	O
of	O
The	O
BBC	O
and	O
Google	B-Application
Inc	I-Application
.	I-Application
</s>
<s>
Some	O
open-source	B-Application
software	I-Application
systems	O
are	O
available	O
,	O
such	O
as	O
:	O
</s>
<s>
Festival	B-Language
Speech	I-Language
Synthesis	I-Language
System	I-Language
which	O
uses	O
diphone-based	O
synthesis	O
,	O
as	O
well	O
as	O
more	O
modern	O
and	O
better-sounding	O
techniques	O
.	O
</s>
<s>
eSpeak	B-Language
which	O
supports	O
a	O
broad	O
range	O
of	O
languages	O
.	O
</s>
<s>
gnuspeech	B-Operating_System
which	O
uses	O
articulatory	B-Application
synthesis	I-Application
from	O
the	B-Operating_System
Free	I-Operating_System
Software	I-Operating_System
Foundation	I-Operating_System
.	O
</s>
<s>
MaryTTS	O
,	O
web	B-Application
based	I-Application
and	O
open	O
source	O
.	O
</s>
<s>
Following	O
the	O
commercial	O
failure	O
of	O
the	O
hardware-based	O
Intellivoice	O
,	O
gaming	O
developers	O
sparingly	O
used	O
software	O
synthesis	O
in	O
later	O
games	O
.	O
</s>
<s>
Earlier	O
systems	O
from	O
Atari	O
,	O
such	O
as	O
the	O
Atari	B-General_Concept
5200	I-General_Concept
(	O
Baseball	O
)	O
and	O
the	O
Atari	B-General_Concept
2600	I-General_Concept
(	O
Quadrun	B-Application
and	O
Open	O
Sesame	O
)	O
,	O
also	O
had	O
games	O
utilizing	O
software	O
synthesis	O
.	O
</s>
<s>
Some	O
e-book	O
readers	O
,	O
such	O
as	O
the	B-General_Concept
Amazon	I-General_Concept
Kindle	I-General_Concept
,	O
Samsung	B-Application
E6	O
,	O
PocketBook	B-Application
eReader	I-Application
Pro	O
,	O
enTourage	B-Application
eDGe	I-Application
,	O
and	O
the	O
Bebook	O
Neo	O
.	O
</s>
<s>
The	O
BBC	B-Device
Micro	I-Device
incorporated	O
the	O
Texas	O
Instruments	O
TMS5220	B-Application
speech	B-Application
synthesis	I-Application
chip	O
,	O
</s>
<s>
Some	O
models	O
of	O
Texas	B-Device
Instruments	I-Device
home	I-Device
computers	I-Device
produced	O
in	O
1979	O
and	O
1981	O
(	O
Texas	B-Device
Instruments	I-Device
TI-99/4	I-Device
and	I-Device
TI-99/4A	I-Device
)	O
were	O
capable	O
of	O
text-to-phoneme	O
synthesis	O
or	O
reciting	O
complete	O
words	O
and	O
phrases	O
(	O
text-to-dictionary	O
)	O
,	O
using	O
a	O
very	O
popular	O
Speech	B-Application
Synthesizer	I-Application
peripheral	O
.	O
</s>
<s>
TI	O
used	O
a	O
proprietary	O
codec	B-General_Concept
to	O
embed	O
complete	O
spoken	O
phrases	O
into	O
applications	B-Application
,	O
primarily	O
video	O
games	O
.	O
</s>
<s>
IBM	O
's	O
OS/2	O
Warp	O
4	O
included	O
VoiceType	O
,	O
a	O
precursor	O
to	O
IBM	B-Application
ViaVoice	I-Application
.	O
</s>
<s>
GPS	O
Navigation	O
units	O
produced	O
by	O
Garmin	O
,	O
Magellan	O
,	O
TomTom	O
and	O
others	O
use	O
speech	B-Application
synthesis	I-Application
for	O
automobile	O
navigation	O
.	O
</s>
<s>
At	O
the	O
2018	O
Conference	O
on	O
Neural	O
Information	O
Processing	O
Systems	O
(	O
NeurIPS	O
)	O
researchers	O
from	O
Google	B-Application
presented	O
the	O
work	O
'	O
Transfer	B-General_Concept
Learning	I-General_Concept
from	O
Speaker	B-Application
Verification	I-Application
to	O
Multispeaker	O
Text-To-Speech	B-Application
Synthesis	O
 '	O
,	O
which	O
transfers	B-General_Concept
learning	I-General_Concept
from	O
speaker	B-Application
verification	I-Application
to	O
achieve	O
text-to-speech	B-Application
synthesis	O
,	O
that	O
can	O
be	O
made	O
to	O
sound	O
almost	O
like	O
anybody	O
from	O
a	O
speech	O
sample	O
of	O
only	O
5	O
seconds	O
.	O
</s>
<s>
Also	O
researchers	O
from	O
Baidu	B-Application
Research	I-Application
presented	O
a	O
voice	O
cloning	O
system	O
with	O
similar	O
aims	O
at	O
the	O
2018	O
NeurIPS	O
conference	O
,	O
though	O
the	O
result	O
is	O
rather	O
unconvincing	O
.	O
</s>
<s>
In	O
March	O
2020	O
,	O
a	O
freeware	B-License
web	B-Application
application	I-Application
called	O
15.ai	B-Algorithm
that	O
generates	O
high-quality	O
voices	O
from	O
an	O
assortment	O
of	O
fictional	O
characters	O
from	O
a	O
variety	O
of	O
media	O
sources	O
was	O
released	O
.	O
</s>
<s>
Initial	O
characters	O
included	O
GLaDOS	O
from	O
Portal	B-Application
,	O
Twilight	O
Sparkle	O
and	O
Fluttershy	O
from	O
the	O
show	O
My	B-Application
Little	I-Application
Pony	I-Application
:	I-Application
Friendship	I-Application
Is	I-Application
Magic	I-Application
,	O
and	O
the	O
Tenth	O
Doctor	O
from	O
Doctor	O
Who	O
.	O
</s>
<s>
A	O
number	O
of	O
markup	O
languages	O
have	O
been	O
established	O
for	O
the	O
rendition	O
of	O
text	O
as	O
speech	O
in	O
an	O
XML-compliant	O
format	O
.	O
</s>
<s>
The	O
most	O
recent	O
is	O
Speech	B-Application
Synthesis	I-Application
Markup	I-Application
Language	I-Application
(	O
SSML	O
)	O
,	O
which	O
became	O
a	O
W3C	O
recommendation	O
in	O
2004	O
.	O
</s>
<s>
Older	O
speech	B-Application
synthesis	I-Application
markup	I-Application
languages	I-Application
include	O
Java	B-Language
Speech	I-Language
Markup	I-Language
Language	I-Language
(	O
JSML	B-Language
)	O
and	O
SABLE	B-Application
.	O
</s>
<s>
Speech	B-Application
synthesis	I-Application
markup	I-Application
languages	I-Application
are	O
distinguished	O
from	O
dialogue	O
markup	O
languages	O
.	O
</s>
<s>
VoiceXML	B-Application
,	O
for	O
example	O
,	O
includes	O
tags	O
related	O
to	O
speech	B-Application
recognition	I-Application
,	O
dialogue	O
management	O
and	O
touchtone	O
dialing	O
,	O
in	O
addition	O
to	O
text-to-speech	B-Application
markup	O
.	O
</s>
<s>
Speech	B-Application
synthesis	I-Application
has	O
long	O
been	O
a	O
vital	O
assistive	O
technology	O
tool	O
and	O
its	O
application	O
in	O
this	O
area	O
is	O
significant	O
and	O
widespread	O
.	O
</s>
<s>
The	O
longest	O
application	O
has	O
been	O
in	O
the	O
use	O
of	O
screen	O
readers	O
for	O
people	O
with	O
visual	O
impairment	O
,	O
but	O
text-to-speech	B-Application
systems	O
are	O
now	O
commonly	O
used	O
by	O
people	O
with	O
dyslexia	O
and	O
other	O
reading	O
disabilities	O
as	O
well	O
as	O
by	O
pre-literate	O
children	O
.	O
</s>
<s>
They	O
are	O
also	O
frequently	O
employed	O
to	O
aid	O
those	O
with	O
severe	O
speech	O
impairment	O
usually	O
through	O
a	O
dedicated	O
voice	B-Application
output	I-Application
communication	I-Application
aid	I-Application
.	O
</s>
<s>
Work	O
to	O
personalize	O
a	O
synthetic	B-Application
voice	I-Application
to	O
better	O
match	O
a	O
person	O
's	O
personality	O
or	O
historical	O
voice	O
is	O
becoming	O
available	O
.	O
</s>
<s>
Speech	B-Application
synthesis	I-Application
techniques	O
are	O
also	O
used	O
in	O
entertainment	O
productions	O
such	O
as	O
games	O
and	O
animations	O
.	O
</s>
<s>
In	O
2007	O
,	O
Animo	O
Limited	O
announced	O
the	O
development	O
of	O
a	O
software	B-Application
application	I-Application
package	O
based	O
on	O
its	O
speech	B-Application
synthesis	I-Application
software	O
FineSpeech	O
,	O
explicitly	O
geared	O
towards	O
customers	O
in	O
the	O
entertainment	O
industries	O
,	O
able	O
to	O
generate	O
narration	O
and	O
lines	O
of	O
dialogue	O
according	O
to	O
user	O
specifications	O
.	O
</s>
<s>
In	O
recent	O
years	O
,	O
text-to-speech	B-Application
for	O
disability	O
and	O
impaired	O
communication	O
aids	O
have	O
become	O
widely	O
available	O
.	O
</s>
<s>
Text-to-speech	B-Application
is	O
also	O
finding	O
new	O
applications	B-Application
;	O
for	O
example	O
,	O
speech	B-Application
synthesis	I-Application
combined	O
with	O
speech	B-Application
recognition	I-Application
allows	O
for	O
interaction	O
with	O
mobile	B-Application
devices	I-Application
via	O
natural	B-Language
language	I-Language
processing	I-Language
interfaces	O
.	O
</s>
<s>
Text-to-speech	B-Application
is	O
also	O
used	O
in	O
second	O
language	O
acquisition	O
.	O
</s>
<s>
Tools	O
,	O
like	O
Elai.io	O
are	O
allowing	O
users	O
to	O
create	O
video	O
content	O
with	O
AI	O
avatars	O
who	O
speak	O
using	O
text-to-speech	B-Application
technology	O
.	O
</s>
<s>
In	O
addition	O
,	O
speech	B-Application
synthesis	I-Application
is	O
a	O
valuable	O
computational	O
aid	O
for	O
the	O
analysis	O
and	O
assessment	O
of	O
speech	O
disorders	O
.	O
</s>
