# MaryTTS - Text to Speech

# This is how to make a marytts server for your own use

# Public tts server config is 4 procs @2.13ghz Xeon with 8GB of RAM

It is currently only using about 3GB with all languages loaded.

I am guessing you can get away with 2gb if your only loading a couple of the voice types.

1. Installed and updated ubuntu 16.04 server.
2. Add webupd8t repo for easy Java install. sudo add-apt-repository ppa:webupd8team/java -Y
3. Update apt list. apt update
4. Insall Java8,unzip,(build-essential,and git) <-Optional. apt install oracle-java8-installer unzip build-essential git -Y
5. Drop root exit
6. Get Marytts installer wget http://mary.dfki.de/download/5.1.2/marytts-installer-5.1.2.zip
7. Extract it unzip marytts-installer-5.1.2.zip

# Now you should have a folder called marytts that you can cd into

1. Change folders with: cd marytts-installer-5.1.2
2. Then give it the intial test run with: ./marytts

# Now you should be able to test the system

Open a web browser and go to http://<ip_of_tts_server>:59125

# Next install your language of choice

At the bottom of this page shows a list of all the languages and voices.

To install all of the voices type the following ~/marytts-installer-5.1.2/marytts install

Or to install a single voice type the following ~/marytts-installer-5.1.2/marytts install:cmu-slt

Replacing the letters after : with your choice.

# -well close to- Oneliner to do a Full LOAD LANGUAGE SUPPORT

Check with wiki and marytts for complete list of language support.

Make sure you are root with the # on the terminal. The script expects to drop sudo, to download and install the tts server.

Otherwise use sudo su

Then this:

add-apt-repository -y ppa:webupd8team/java && apt update && apt install -y oracle-java8-installer unzip build-essential git && exit

wget http://mary.dfki.de/download/5.1.2/marytts-installer-5.1.2.zip && unzip marytts-installer-5.1.2.zip && cd marytts-installer-5.1.2 && ./marytts install

# Setup MarryTTS in ~/.config/naomi/configs/profile.yml

⚠️ Don't forget to add a voice and a valid IP address & port

Now you need to update your ~/.jasper/profile.yml:

# Text To Speech Config default is espeak-tts
# uncomment to use
tts_engine: mary-tts

mary-tts:
  server: 'yourserverip'
  port: '59125'
  language: 'en_US'
  voice: 'cmu-slt-hsmm'

Better-Secured(https) to the server:

Coming Soon (code base update required).

# Voices for Marry TTS

Supported Languages BOLD ARE CHOICES FOR profile.yml:

# English

• cmu-bdl-hsmm :A male US English hidden semi-Markov model voice, built from recordings provided by Carnegie Mellon University

• cmu-slt :A female US English unit selection voice, built from recordings provided by Carnegie Mellon University

• cmu-slt-hsmm :A female US English hidden semi-Markov model voice, built from recordings provided by Carnegie Mellon University

• cmu-rms-hsmm :A male US English hidden semi-Markov model voice, built from recordings provided by Carnegie Mellon University

• dfki-obadiah :A male British English expressive unit selection voice: Gloomy Obadiah

• dfki-obadiah-hsmm :A male British English hidden semi-Markov model voice

• dfki-poppy :A female British English expressive unit selection voice: Cheerful Poppy

• dfki-poppy-hsmm :A female British English hidden semi-Markov model voice

• dfki-prudence :A female British English expressive unit selection voice: Pragmatic Prudence

• dfki-prudence-hsmm :A female British English hidden semi-Markov model voice

• dfki-spike :A male British English expressive unit selection voice: Aggressive Spike

• dfki-spike-hsmm :A male British English hidden semi-Markov model voice

# French

• enst-camille :A female French unit selection voice, built at Télécom ParisTech (ENST) using data recorded by Camille Dianoux

• enst-camille-hsmm :A female French hidden semi-Markov model voice, built at Télécom ParisTech (ENST) using data recorded by Camille Dianoux

• enst-dennys-hsmm :A male Québécois French hidden semi-Markov model voice, built at Télécom ParisTech (ENST)

• upmc-jessica :A female French unit selection voice, built at ISIR (UPMC) using data recorded by Jessica Durand

• upmc-jessica-hsmm :A female French hidden semi-Markov model voice, built at ISIR (UPMC) using data recorded by Jessica Durand

• upmc-pierre :A male French unit selection voice, built at ISIR (UPMC) using data recorded by Pierre Chauvin

• upmc-pierre-hsmm :A male French hidden semi-Markov model voice, built at ISIR (UPMC) using data recorded by Pierre Chauvin

# German

• dfki-pavoque-neutral :A male German unit selection voice

• dfki-pavoque-neutral-hsmm :A male German hidden semi-Markov model voice

• dfki-pavoque-styles :A male German unit selection voice with expressive styles "happy", "sad", "angry", and "poker"

• bits1-hsmm :A female German hidden semi-Markov model voice, built from voice recordings provided by the BITS project at the Bavarian Archive of Speech Signals

• bits3 :A male German unit selection voice, built from voice recordings provided by the BITS project at the Bavarian Archive of Speech Signals

• bits3-hsmm :A male German hidden semi-Markov model voice, built from voice recordings provided by the BITS project at the Bavarian Archive of Speech Signals

# Italian

• istc-lucia-hsmm :Italian female Hidden semi-Markov model voice kindly made available by Fabio Tesser

# Turkish

• dfki-ot :A male Turkish unit selection voice

• dfki-ot-hsmm :A male Turkish hidden semi-Markov model voice

# Russian

• voxforge-ru-nsh :Russian male voice kindly made available by Nickolay V. Shmyrev

# Telugu

• cmu-nk :A female Telugu unit selection voice built from voice recordings provided by IIIT Hyderabad and Carnegie Mellon University

• cmu-nk-hsmm :A female Telugu hidden semi-Markov model voice built from voice recordings provided by IIIT Hyderabad and Carnegie Mellon University