Rockbox mail archiveSubject: Re: Voice file generation for various languages
Re: Voice file generation for various languages
From: Rocker <rocker_at_shaw.ca>
Date: Thu, 29 Apr 2004 19:46:36 -0600
Nice work man!
----- Original Message -----
From: "Jens Arnold" <arnold-j_at_t-online.de>
Sent: Thursday, April 29, 2004 5:26 PM
Subject: Voice file generation for various languages
I have prepared a VBScript to generate voice files for the
Rockbox Voice UI for various languages. It can utilize both
SAPI5 and SAPI4 voices, so it is possible to generate voice
files for quite a number of languages.It also features internal
lists to tweak pronunciation without the need to introduce
deliberate spelling errors (which would be engine dependent)
into the .lang files.
Languages/voices I've already done (These are available at
http://arnold-j.bei.t-online.de/Rockbox/ and I will try to keep
them in sync with the corresponding .lang files):
German, AT&T Natural Voices Klara (deutsch_Klara.zip)
US English, AT&T Natural Voices Crystal (english_Crystal.zip)
US English, AT&T Natural Voices Mike (english_Mike.zip)
US English, Microsoft Mary (english_MSMary.zip)
US English, Microsoft Mike (english_MSMike.zip)
US English, Microsoft Sam (english_MSSam.zip)
These files are generated with tweaked pronunciation.
Languages/voices I should be able to prepare with the freely
available L&H TTS3000 engines provided that the voices entries
in the corresponding .lang file are created by someone with a
knowledge of the language first:
French (Veronique, Pierre), Dutch (Karen, Hans),
Italian (Barbara, Stefano), Spanish (Carmen, Julio),
Brazilian Portuguese (Juliana, Alexandre)
Beware that the L&H TTS3000 voices don't sound very pleasant.
That's why I didn't list the German and British english voices,
since these would be by far inferior to the already available
AT&T and Microsoft voices.
It would also be a raw version - since I don't know these
languages, I cannot tweak pronunciation.
Voice file generation for SAPI5 voices (AT&T, Microsoft) works
really fast (~40 sec for a complete voice file on my system),
unfortunately using SAPI4 voices (L&H) is rather slow, because
it has to be done in realtime (~15 min for a complete voice file).
At least it runs unattended.
If someone wants to try the script for himself, it can be found
at http://arnold-j.bei.t-online.de/Rockbox/ as well, however, it
is still not perfect and there is no description yet. Additionally
you will need WavTrim.exe (by Jörg, modified by me),
VoiceFont.exe (also by Jörg) and Lame.exe, all of them either in
the path or in the same directory as the script.
Received on 2004-04-30