Text to Speech


MaryTTS is a java-based open-source text to speech system.
It was developed by the German Research Center for Artificial Intelligence and the University of Saarland.
In our case we use the MaryTTS-Installer to both install new languages and run the server,
because we were unable to install new languages with the Installer-GUI due to our missing video output.

Install MaryTTS


Download the MaryTTS-Installer and unzip it.
Go to the directory, where you downloaded it and install a language:

    cd marytts-installer-5.2/
    ./marytts install voice-bits3-hsmm

bits3-hsmm is a german male voice, you can try any voices here.

To run your server you just need to type ./marytts in the same directory.
The server will be available at http://localhost:59125/.

Run MaryTTS as a service

We created a service to start the MaryTTS server at boot up.
To do so, type sudo nano /etc/systemd/system/marytts.service in the console and copy the following lines:

    # Make sure to change ExecStart, WorkingDirectory
    # and Environment to the path, where you installed marytts


Save and close the file.
Enable and start the service by typing:

    sudo systemctl enable marytts.service
    sudo systemctl start marytts.service

Use MaryTTS in Rhasspy

For MaryTTS you need to have OPENSSL_1_1_1 installed.
To see how to do this, look at our troubleshooting-section for OpenSSL.

Update your profile

Go to your Rhasspy-profile or the webinterface and add the following lines:

  "text_to_speech": {
      "marytts": {
          "locale": "de",
          "voice": "bits3-hsmm"
      "system": "marytts"

Make sure to change "locale" and "voice" to any voice you installed.

Test MaryTTS

Now you should be able to test MaryTTS by typing any word or sentence in the input-field of the webinterface.
By pressing the “Speak”-button you should be able to hear the spoken text through your selected “Audio Playing”-service, in our case aplay (3.5mm-headphonejack).


Call TTS from Node-Red

To see how we use Node-Red to interact with MaryTTS, click here.


Universität des Saarlands

What’s next?

If you do not want to use Rhasspy’s own dialogue manager you can use HermesMQTT.