Description

Mycroft TTS plugin for Cotovia TTS

About Cotovia

Cotovia is a unit-selection text-to-speech system, i.e., generates the synthetic speech signal as a concatenation of prerecorded segments. Cotovia determines the sequence of sounds, its intonation and duration from the input text. Actually several intonation contours are considered in parallel and for each one a sequence of speech units is selected. The final intonation contours is selected according to the suitability of the sequence of speech units found.

Voices

At this time two Galician voices are available at sourceforge. These two voices are also suitable for Spanish, with little distortion, since the phonemes of Spanish may be considered a subset of Galician phonemes. The nicknames of the two speakers are Iago and Sabela. Nevertheless you will find three voices available for Cotovia:

• iago : Galician male speaker, duration of the recording: 80 minutes. It was Cotovia's first voice and the quality is limited by the reduced recording time.

• sabela-large : Galician female speaker, duration of the recording: 14.5 hours. Use this voice if you want to obtain maximum quality speech and you are not too worried about execution time.

• sabela: default speaker. A subset of Sabela's recordings (about 4 hours). A good compromise between quality and execution time.

Install

Download and install Cotovia

In order to run Cotovia, you will need to install the following packages:

cotovia_0.5_amd64.deb   ---  Cotovia executable
cotovia-lang-gl_0.5_all.deb --- Galician linguistic data
cotovia-lang-es_0.5_all.deb --- Spanish linguistic data

Additionally you need to install at least one voice

cotovia-voice-iago_0.5_all.deb
cotovia-voice-sabela-large_0.5_all.deb
cotovia-voice-sabela_0.5_all.deb

Then install the plugin

pip install ovos-tts-plugin-cotovia

Configuration

  "tts": {
    "module": "ovos-tts-plugin-cotovia",
    "ovos-tts-plugin-cotovia": {
      "voice": "iago"
    }
  }
 

Advanced config

Additional configuration params are available

• lang can be gl for galician or es
• voice cab be iago or sabela
• pitch_scale_factor can be used to change pitch (default 100)
• time_scale_factor can be used to change speed (default 100)
• bin can be used to set a path to the executable (default /usr/bin/cotovia)

  "tts": {
    "module": "ovos-tts-plugin-cotovia",
    "ovos-tts-plugin-cotovia": {
      "voice": "sabela",
      "lang": "es",
      "pitch_scale_factor": 80,
      "time_scale_factor": 80,
      "bin": "/usr/bin/cotovia"
    }
   }