Discussion list for building synthetic voices using Festvox tools ()

wolf-N+oXqPB6o/YYGKtr5sm4Aw@public.gmane.org | 7 Apr 2008 13:18

german language speaker for festvox/festival?

hello.

im quite new to festival/festvox, so i hope i dont annoy with redundant
questions. i search the net if there is any german language speaker
support for text to speech yet. apart from some older references that
this is worked on, i couldnt find anything.

does anybody know if there are people on this issue or if there is
even something usable around allready? and if not at all, is there a
tutorial that would help a non-festvox-expert to adjust english in a
way germn text would sound at least somehow german?

wolf.

Nickolay V. Shmyrev | 7 Apr 2008 18:45

Picon

Re: german language speaker for festvox/festival?

07.04.08, 15:18, "wolf@..." <wolf@...>:

> hello.
> im quite new to festival/festvox, so i hope i dont annoy with redundant
> questions. i search the net if there is any german language speaker
> support for text to speech yet. apart from some older references that
> this is worked on, i couldnt find anything.
> does anybody know if there are people on this issue or if there is
> even something usable around allready? and if not at all, is there a
> tutorial that would help a non-festvox-expert to adjust english in a
> way germn text would sound at least somehow german?
> wolf. 

Nothing completely free, but you can try 

http://www.ims.uni-stuttgart.de/phonetik/synthesis/

Also if you aren't bound to festival, try OpenMARY, they provide rather decent support for German.

http://mary.dfki.de/Download

If you'd like to step on and create free German voice, you are welcome. The data
is available, someone just need to preprocess it.

wolf-N+oXqPB6o/YYGKtr5sm4Aw@public.gmane.org | 8 Apr 2008 11:32

Re: german language speaker for festvox/festival?

On Mon, 07 Apr 2008 20:45:12 +0400
"Nickolay V. Shmyrev" <nshmyrev@...> wrote:

> 
> Nothing completely free, but you can try 
> 
> http://www.ims.uni-stuttgart.de/phonetik/synthesis/
> 
> Also if you aren't bound to festival, try OpenMARY, they provide
> rather decent support for German.
> 
> http://mary.dfki.de/Download
> 

thx, unfortunately im working on a mipsel system, so as i will have to
compile and adjust myself, i will need a free/openb solution.
also jave is no option.

> If you'd like to step on and create free German voice, you are
> welcome. The data is available, someone just need to preprocess it.

the problem is that i dont even know what preprocessing means :( here.
if someone on this list can take my hand and help me i can spend some
of time on it.

w.

Nickolay V. Shmyrev | 8 Apr 2008 13:06

Picon

Re: german language speaker for festvox/festival

Well, first of all let's define your requirements. Festival is actually
sometimes more resource consuming than Java so probably it's not an
option for you as well.

What is the resources of your system - how many memory do you have, what
is the CPU performance?

What is the size of the vocabulary you will process? Many optimizations
depend on that.

What is the quality of synthesis you need? Current methods allows you to
build high-quality voice but it will use more than 100Mb of memory and
will consume a lot of resources. Unit selection is probably not an
option for you. Modern HTS voices are smaller, but MLSA also requires
significant computational resources. Probably you just need a diphone
voice. Btw, did you consider espeak (http://espeak.sourceforge.net).
It's quality not very natural but it's a good choice for resource-
constrained system.

wolf-N+oXqPB6o/YYGKtr5sm4Aw@public.gmane.org | 8 Apr 2008 13:51

Re: german language speaker for festvox/festival

On Tue, 08 Apr 2008 15:06:38 +0400
"Nickolay V. Shmyrev" <nshmyrev@...> wrote:

> Well, first of all let's define your requirements. Festival is
> actually sometimes more resource consuming than Java so probably it's
> not an option for you as well.

> What is the resources of your system - how many memory do you have,
> what is the CPU performance?

my specs are 233Hhz mipsel (Boadcom) processors with variable amount of
RAM between 32 and 64MB. im running openwrt as OS, but running festival
inside a debian changeroot to satisfy all library dependencies.

i tried java before and found it not suitable to run ...

> What is the size of the vocabulary you will process? Many
> optimizations depend on that.

ok, i admit i know very little about that.

> What is the quality of synthesis you need? Current methods allows you
> to build high-quality voice but it will use more than 100Mb of memory
> and will consume a lot of resources. Unit selection is probably not an
> option for you. Modern HTS voices are smaller, but MLSA also requires
> significant computational resources. Probably you just need a diphone

the sound quality does not necessarily 
have to be "natural", of course you always want it "as best as
possible".

(Continue reading)

Seann Clark | 10 Apr 2008 16:54

building a different sounding voice in an established language

All,

    This is my first post, and didn't find anything on looking for the 
information. As a matter of fact, aside from the building voices how to 
(a bit complex to read, but very informative and useful) I haven't found 
anything to get what I would want. I have a very small-ish project, of 
using MisterHouse, and like programs, to talk using festival. The 
default voices are great for testing, but I am looking at using a 
specific voice print to speak using festival in English. Actually I have 
two ideas,  one is to extract the voice samples of the voice I want 
(voice acted out, out of a series of DVD's I have) to get the voice 
sound I desire (since I have a hard time finding a person in real life 
to help me with the voice sound I desire.  the other idea is to use a 
voice from a different language (French, German, Japanese, something 
like that) and use it on an English language speaking Festival.

    I don't know if it is possible to use existing portions of pre-built 
voices and extend them, and move the speech patterns and diphones over, 
or if it would be better to just build everything from scratch (which I 
have never done, as of yet) and do it that way.

    With the building of the voices, is it possible to get smooth 
natural sounding phrases on the fly? As I have said, and/or alluded to, 
I am a beginner with this, and I am used to items like Nuance speech 
engines, and so forth.

~Seann

Nickolay V. Shmyrev | 11 Apr 2008 05:44

Picon

Re: building a different sounding voice in an established language


В Чтв, 10/04/2008 в 09:54 -0500, Seann Clark пишет:
> All,
> 
>     This is my first post, and didn't find anything on looking for the 
> information. A

Hi Seann, you are welcome

> s a matter of fact, aside from the building voices how to 
> (a bit complex to read, but very informative and useful) I haven't found 
> anything to get what I would want. I have a very small-ish project, of 
> using MisterHouse, and like programs, to talk using festival. The 
> default voices are great for testing, but I am looking at using a 
> specific voice print to speak using festival in English. Actually I have 
> two ideas,  one is to extract the voice samples of the voice I want 
> (voice acted out, out of a series of DVD's I have) to get the voice 
> sound I desire (since I have a hard time finding a person in real life 
> to help me with the voice sound I desire.  the other idea is to use a 
> voice from a different language (French, German, Japanese, something 
> like that) and use it on an English language speaking Festival.

Looks like an option for you. If you can easily extract your voice from
the movie, it will work though will not give you best results. Tools
aren't oriented on compressed speech. Probably it's easier to find voice
talent and record him.

>     I don't know if it is possible to use existing portions of pre-built 
> voices and extend them, and move the speech patterns and diphones over, 
> or if it would be better to just build everything from scratch (which I

(Continue reading)

Seann Clark | 11 Apr 2008 15:39

Re: building a different sounding voice in an established language

Nickolay V. Shmyrev wrote:
> В Чтв, 10/04/2008 в 09:54 -0500, Seann Clark пишет:
>   
>> All,
>>
>>     This is my first post, and didn't find anything on looking for the 
>> information. A
>>     
>
> Hi Seann, you are welcome
>
>   
>> s a matter of fact, aside from the building voices how to 
>> (a bit complex to read, but very informative and useful) I haven't found 
>> anything to get what I would want. I have a very small-ish project, of 
>> using MisterHouse, and like programs, to talk using festival. The 
>> default voices are great for testing, but I am looking at using a 
>> specific voice print to speak using festival in English. Actually I have 
>> two ideas,  one is to extract the voice samples of the voice I want 
>> (voice acted out, out of a series of DVD's I have) to get the voice 
>> sound I desire (since I have a hard time finding a person in real life 
>> to help me with the voice sound I desire.  the other idea is to use a 
>> voice from a different language (French, German, Japanese, something 
>> like that) and use it on an English language speaking Festival.
>>     
>
> Looks like an option for you. If you can easily extract your voice from
> the movie, it will work though will not give you best results. Tools
> aren't oriented on compressed speech. Probably it's easier to find voice
> talent and record him.

(Continue reading)

kazu.hodota | 12 Apr 2008 04:43

Picon

Re: building a different sounding voice in an established language

Hello Seann Clark,

Just for your information,  we have some Japanese speaking software on 
Linux.

1. Galatea Project :

"Galatea" is a project for providing an open-source, license-free 
software toolkit for building anthropomorphic spoken dialogue agents.....

http://hil.t.u-tokyo.ac.jp/~galatea/index.html

2. TTSynth  Text To Speech (TTS) software :

Available Modules Include:
IBM's incomparable IBM TTS text to speech (formerly called ViaVoice)
Gnome-Speech patched for use with LSR and Orca
The TTSynth Speakup Bridge for screen reading with Speakup
ttsynth_say binary for the CLI or for use in scripts
Also supported by Emacspeak KTTS, and Speech Dispatcher

http://ttsynth.com/

Yours,

Kazu Hodota

Seann Clark さんは書きました:
> Nickolay V. Shmyrev wrote:
>> В Чтв, 10/04/2008 в 09:54 -0500, Seann Clark пишет:

(Continue reading)

kazu.hodota | 12 Apr 2008 05:08

Picon

Re: building a different sounding voice in an established language

Hello Seann Clark,

Just for your information,  we have some Japanese speaking software on 
Linux.

1. Galatea Project :

"Galatea" is a project for providing an open-source, license-free 
software toolkit for building anthropomorphic spoken dialogue agents.....

http://hil.t.u-tokyo.ac.jp/~galatea/index.html

2. TTSynth  Text To Speech (TTS) software :

Available Modules Include:
IBM's incomparable IBM TTS text to speech (formerly called ViaVoice)
Gnome-Speech patched for use with LSR and Orca
The TTSynth Speakup Bridge for screen reading with Speakup
ttsynth_say binary for the CLI or for use in scripts
Also supported by Emacspeak KTTS, and Speech Dispatcher

http://ttsynth.com/

Yours,

Kazu Hodota

Seann Clark さんは書きました:
> Nickolay V. Shmyrev wrote:
>> В Чтв, 10/04/2008 в 09:54 -0500, Seann Clark пишет:

(Continue reading)

Group

gmane.science.tts.festvox.

Project Web Page

Discussion list for building synthetic voices using Festvox tools ()

Search Archive

Page

1 | 2

Articles in period: 13

April 2008

Language

Change language

Options

Current view: Threads only / Showing only 20 lines / Not hiding cited text.
Change to All messages, whole messages, or hide cited text.

Post a message
NNTP Newsgroup
Classic Gmane web interface
XML

XML

RSS Feed
List Information

About Gmane

Recent entries

bsv using HMM (20 May 2013)
building synthetic voice using festvox tools and speech tools (10 May 2013)
building synthetic voice using festvox tools and speech tools (8 May 2013)
building synthetic voice using festvox tools and speech tools (7 May 2013)
Únete a mi red en LinkedIn (12 Mar 2013)
Multisyn Voice (26 Feb 2013)
hi festvox (31 Jan 2013)
"Discarded due to NAN problem" error (9 Jan 2013)
Error : "core dumped" or "'std::bad_alloc" cause of big clunits voice (1 Jan 2013)
Stay in touch with me through LinkedIn (22 Dec 2012)
Too long to finish building spectral model (do_clustergen cluster) (17 Dec 2012)
festvox (13 Dec 2012)
Error when executing ./bin/do_clustergen cluster (10 Dec 2012)
festvox (4 Dec 2012)
Error when create cmu_us_slt_arctic_mcep.tree (1 Dec 2012)
Error when create cmu_us_slt_arctic_mcep.tree (1 Dec 2012)
(unknown) (30 Nov 2012)
festvox (30 Nov 2012)
festvox (25 Oct 2012)
text to speech system (8 Oct 2012)
text to speech system (8 Oct 2012)

Archives

10	May 2013
1	March 2013
2	February 2013
3	January 2013
9	December 2012
3	November 2012
9	October 2012
7	September 2012
1	June 2012
1	February 2012
2	December 2011
5	November 2011
2	October 2011
5	August 2011
1	July 2011
11	June 2011
2	May 2011
2	April 2011
5	January 2011
2	December 2010
6	November 2010
8	August 2010
3	July 2010
8	June 2010
2	May 2010
6	April 2010
11	March 2010
8	January 2010
11	December 2009
13	November 2009
4	October 2009
12	September 2009
9	August 2009
5	July 2009
21	June 2009
13	May 2009
25	April 2009
17	March 2009
1	February 2009
1	January 2009
2	December 2008
8	November 2008
5	October 2008
21	September 2008
6	July 2008
9	June 2008
13	April 2008
4	February 2008
7	January 2008
35	December 2007
31	November 2007
4	October 2007
25	September 2007
8	July 2007

Design

Zawodny | Right Menu | Left Menu

Your Own Design

Paste the URL of your CSS below. Download CSS template