| I hope to create a proper man page and add it fairly soon. It is clearly necessary.
You may be interested to listen to:
and
The last item provides a comparison of gnuspeech and natural speech, repeated three times (the examples are all repeated three times, in fact).
All the artificial speech samples were created by gnuspeech from normally punctuated plain text with no adjustments. The first set of samples were created from the command line input option of gnuspeechsa without any change in the initial options, as described in Appendix D of the Gnuspeech manual. The second and third items were generated from Monet using the gnuspeech program suite that runs on Mac OS X.
I hope this helps.
david
-------- David Hill -------- Simplicity, patience, compassion. These three are your greatest treasures (Tao Te Ching #67) ---------
On Oct 26, 2015, at 10:23 50AM, Advrk Aplmrkt wrote: Yes, it works now. But after hearing the defaults it sounds like lots of tweaking to the configuration is need to make the voice sound more real? Right now it sounds very funny, and not nearly as good as, say, Siri. Let me know if you have any tips for optimising the voice. Thanks for the instructions! On 23/10/2015, Marcelo Y. Matuda < address@hidden> wrote: Hi,
To synthesize a short text:
./gnuspeech_sa -c ../gnuspeechsa-0.1.5/data/en/ -p param.txt \
-o out.wav "This is a test."
To synthesize from a text file:
./gnuspeech_sa -c ../gnuspeechsa-0.1.5/data/en/ -i input.txt \
-p param.txt -o out.wav
To test another voice, change the parameter voice_name in
../gnuspeechsa-0.1.5/data/en/trm_control_model.config.
You can change other parameters in that file, for example tempo and
pitch_offset.
Try this in voice_male.config:
vocal_tract_length = 22.0
reference_glottal_pitch = -20.0
Regards,
Marcelo
On 10/21/2015 01:28 PM, David Hill wrote:
As a first step, check out what is written in Appendix D of the Monet
Manual concerning producing speech from gnuspeechsa.
gnuspeechsa is a command-line program, and you should be able to produce
speech by entering an appropriate command, which will result in a sound
file in the location you specify which you can play. For example, if you
are in the directory into which gnuspeechsa has been compiled, the
command would be:
./gnuspeechsa -v -c ../../data/en -p trm_param_file.txt -o
do-you-happen.wav "Do you happen to know by chance what time it is?"
(Note that the manual has unfortunately got a misprint for this command
example, by inserting a "\_" in gnuspeechsa.)
There's a recording on YouTube of the resulting speech output that is
produced, repeated three times at:
https://www.youtube.com/watch?v=NUrODiSkWk8&feature=youtu.be
I have also passed a copy of your email to the contributor responsible
for gnuspeechsa. He has had to drop participation in GNU project for
now, in order to take time out for other things, but may be able to make
some helpful comments which I can pass on to you.
In GNU/Linux applications users may use Gnuspeechsa for output from an
application by using Speech-dispatcher:
YELAVICH, LUKE , JAN BUCHAL, TOMAS CERHA, HYNEK HANKE, MILAN ZAMAZAL,
C.M. BRANNON, WILLIAM HUBBS, ANDREI KHOLODNY (undated). Speech
Dispatcher, http://devel.freebsoft.org/speechd, accessed 2015-07-24.
Yes, the release manuals are quite technical, but the technology
underlying synthetic speech is not simple. You can learn a lot by
studying the background matyerial in the manuals, as well as some of the
papers that are linked.
Happy hacking!
david
--------
David Hill
address@hidden <mailto:address@hidden>
http://www.gnu.org/software/gnuspeech/
http://savannah.gnu.org/projects/gnuspeech
https://savannah.gnu.org/users/davidhill
--------
Simplicity, patience, compassion. These three are your greatest
treasures (Tao Te Ching #67)
---------
On Oct 21, 2015, at 5:58 49AM, Advrk Aplmrkt wrote:
Hello,
I just read about the latest release of Gnuspeech, and am eager to
take it for a spin. I'm not a speech synthesis researcher, I just like
to easily do text to speech with Free Software.
I found the Gnuspeech Monet Manual 0.9 and the Gnuspeech TRAcT Manual
0.9, but they look very dense and technical. I've got gnuspeech-sa
compiled and installed on my Fedora Linux system, but I have no idea
how to run it by feeding it either a .txt file or just direct input of
text from the command line. It seems to ask for some sort of
configuration file?
What should I do to just get it to start speaking text? Is there a
quickstart guide or tutorial? Or maybe some preset "voices" I can pick
from? Thanks!!
|