Needybot Speech

Open Source speech client for Needybot.

This package uses Ivona for raw speech generation, a free developer account is required.

Getting started

All apt-get and python (pip) dependencies are installed via a command in CMakeLists.txt. Currently, the dependencies required are:

• ladspa-sdk
• libsox-fmt-all
• python-pygame
• sox
• tap-plugins

And in requirements.txt:

• pydash
• pyvona
• requests

This project expects you will be working within a catkin workspace. Assuming you have a catkin workspace set up in your home directory, you can get up and running by doing the following:

$ cd ~/catkin_ws/src
$ git clone https://github.com/needybot/needybot-speech.git
$ cd ~/catkin_ws
$ catkin_make

If all does not go well, please read our contributing guidelines and create an issue.

Environment Variables and ROS Parameters

ENV

The following Ivona environment variables must be set for this package to work:

• IVONA_ACCESS_KEY
• IVONA_SECRET_KEY

ROSPARAM

• /needybot/speech/voice/name: string ['Justin']
  • The voice identifier for Ivona
• /needybot/speech/voice/speech_rate: string ['medium']
  • Uses standard SSML prosody rate labels
    1. x-slow
    2. slow
    3. medium (default)
    4. fast
    5. x-fast
    6. default (medium)
• /needybot/speech/voice/codec: string ['ogg']
  • The audio encoding format
    1. mp3
    2. mp4
    3. ogg
    4. speechmark
• /needybot/speech/voice/cache_dir: string ['/path/to/needybot_speech/cache']
  • The path at which your cache manifest and audio files will live.
• /needybot/speech/voice/clear_cache: boolean [false]
  • Flag signaling that you would like to completely clear your speech cache.
• /needybot/speech/voice/warm_cache: boolean [true]
  • Flag signaling that you would like to prewarm the speech cache.

Adding Dialog

By default, all dialog is stored in param/dialog_manifest.yml, however, you may override this when including launch/needybot_speech.launch in your own file with the <arg name="manifest" value="/abs/path/to/my_dialog_manifest.yml" /> tag. All dialog will be loaded into the rosparam server under the /needybot/speech/dialog namespace. For example, your manifest might look like this:

just_hello:
  template: "Hello!"
introduction:
  template: "Hello world! My name is ${name} and my favorite number is ${favorite_number}."
  params:
    name: "Needybot"
    favorite_number: "pi"

In the first entry labeled just_hello, Needybot will be fed the template and speak it normally with no substitutions (substitutions are optional). In the second entry labeled introduction, we see a template with placeholder identifiers. We use the Template class provided by the string module in python's standard library. Beneath the template is a dictionary of default parameters, should you not provide any at run time.

TODO: action server example

Contributing

See our contribution guidelines here.

Testing

WARNING: Running package tests will clean and warm your dialog cache directories. Keep this in mind when working with production code and be sure to backup your data when necessary.

This package uses the rostest command for testing, to run tests:

$ rostest needybot_speech needybot_speech.test

License

This project is release under Apache 2.0. Please see the LICENSE file for more details.