| ROS Resources: Documentation | Support | Discussion Forum | Index | Service Status | ros @ Robotics Stack Exchange |
Hi all,
I'm trying to make our pr2 speak. However the (sound_play node + say.py script) gives a deep and slow voice hardly understandable. I changed voices in festival, but sound_play doesn't take it into account. How to make sound_play use the voice set as default in festival ?
Guido
Hi Guido,
I submitted a ticket for the "deep slow voice" issue about 6 weeks ago. In the meantime, the only way I figured out how to change the voice is as follows. (This applies to Ubuntu 10.04 with Diamondback debian packages and Festival debian packages.)
$ cd /usr/share/festival
$ sudo pico voices.scm
(Or use whatever editor you like instead of pico.)
Once in the editor, search for "kal_diphone" (without the quotes). Change this to "don_diphone" (again without the quotes).
Save changes and exit the editor.
You should now hear a better voice when using sound_play node + say.py.
--patrick
I solved this by editing /opt/ros/fuerte/stacks/audio_common/sound_play/scripts/say.py
Do the following:
roscd sound_play
sudo pico scripts/say.py
In this file you need to change the following line:
voice = 'voice_kal_diphone'
I changed it to (assuming you have other voices installed in festival):
voice = 'voice_nitech_us_rms_arctic_hts'
Changing the frequency used by text2wav helps, but your best bet is to check out joq's answer to another question.
Use rosrun sound_play say.py "Text to say" voice_you_want
In C++
sound_play::SoundClient sc; sc.say("Phrase to say.","voice_you_want");
If you write "" in voice_you_want, you play the festival default voice
I took the example from isura:
roscd sound_play
sudo pico scripts/say.py
sound_play use the voice set as default in festival when you leave voice blank:
voice = ''
it works for me with a german Mbrola voice with the special ims-festival (german language) from the university of stuttgart
Asked: 2011-06-10 00:58:42 -0500
Seen: 1,791 times
Last updated: Jan 13 '13
ROS Answers is licensed under Creative Commons Attribution 3.0 Content on this site is licensed under a Creative Commons Attribution Share Alike 3.0 license.