MBROLA Hi-Quality Free Speech Synthesis

MBROLA Hi-Quality Free Speech Synthesis

Post by Vincent Pag » Sat, 29 Mar 1997 04:00:00



M     M  BBBB    RRRR      OOO    L        A
MM   MM  B   B   R   R    O   O   L       A A
M M M M  B   B   R   R   O     O  L      A   A
M  M  M  BBBB    RRR     O     O  L     AAAAAAA
M     M  B   B   R  R    O     O  L     A     A
M     M  B    B  R   R    O   O   L     A     A
M     M  BBBBB   R    R    OOO    LLLLL A     A

(MBROLA should read similar to 'umbrella')

HIGH-QUALITY DIPHONE-BASED SPEECH SYNTHESIZERS AVAILABLE FOR FREE
                    IN THE CONTEXT OF THE
               MBROLA SPEECH SYNTHESIS PROJECT
------------------------------------------------------------------

Nearly one year ago, we announced here that the TCTS Lab of the
Faculte Polytechnique de Mons (Belgium) had launched an ambitious
speech synthesis project, the aim of which was to obtain a set of
speech synthesizers for as many languages as possible, FREE FOR USE IN
NON-COMMERCIAL, NON-MILITARY APPLICATIONS.

We started the MBROLA project with a single diphone database for
French, and we called for international collaborations for the
development of free databases in other languages.

Thanks to a number of laboratories and individuals who have understood
the importance of free speech synthesis for non-commercial purposes,
the project is a complete succes.  

As a result, we are proud to announce the availability of seven
diphone databases :

br1: a Brasilian Portuguese male voice provided by Denis Costa, Sao Paolo
de1: a German female voice provided by Fred Englert, University of
Stuttgart.
en1: a British English male voice provided by Alan Black, CSTR, Edimburgh.
fr1: a French male voice provided by Thierry Dutoit, FPMs Mons Belgium.
nl1: a Dutch male voice provided by Arthur Dirksen and Ludmila Menert.
ro1: a Romanian male database provided by Marian Boldea, University of
     Timisoara.
es1: a spanish male voice developed by university autonoma of Barcelona

Central to this project, is MBROLA v2.03, a speech synthesizer
based on the concatenation of diphones.

THIS SYNTHESIZER IS AVAILABLE FOR FREE.
Executable code is provided for PC/DOS, PC/LINUX, SUN4 and 5, HPUX,
VAX/VMS, DECALPHA/VMS, PC/SOLARIS2.4, , SGI INDY/IRIX, NEXT/NEXTSTEP,
IBMRS6000.

MBROLA takes a list of phonemes as input, together with prosodic
information (duration of phonemes and a piecewise linear description
of pitch : no prosodic model is assumed), and produces speech samples
on 16 bits (linear), at the sampling frequency of the diphone
database. (It is therefore NOT a Text-To-Speech (TTS) synthesizer,
since it does not accept raw text as input. Pointers to freely
available text-to-phonetics systems interfaced with MBROLA are
provided, however.

THE ULTIMATE GOAL OF THIS PROJECT IS TO BOOST UP ACADEMIC RESEARCH ON
SPEECH SYNTHESIS, AND PARTICULALRY ON PROSODY GENERATION, KNOWN AS
ONE OF THE BIGGEST CHALLENGES TAKEN UP BY TEXT-TO-SPEECH SYNTHESIZERS
FOR THE YEARS TO COME.

More details can be found at the MBROLA project homepage :

   http://tcts.fpms.ac.be/synthesis

(includes mirror sites in France Switzerland, and US)

------------------------------------------------------------------
The mbrola team

Faculte Polytechnique de Mons, TCTS Lab,
31, bvd Dolez, B-7000 Mons, Belgium.
tel   : /32/65/374133
fax   : /32/65/374129

questions on the installation of software and databases,
user registration