Software for the Text-to-Speech tutorial at SLTU 2016. Examples of how to use some of these tools can be found in the Lab Session section of the slide deck for the tutorial.

The Dockerfile here is used to automatically build the image mjansche/tts-tutorial and as a basis for the larger image mjansche/tts-tutorial-sltu2016 on Docker Hub.

It is based on Ubuntu 16.04 LTS ("Xenial Xerus") and contains the following software:

• Festival 2.4
• Edinburgh Speech Tools 2.4
• FestVox 2.7
• SPTK 3.6
• OpenFst 1.5.2
• Thrax 1.2.2
• Pynini 0.9
• Sparrowhawk 0.1
• Protocol Buffers
• RE2
• NCHLT-inlang Afrikaans Pronunciation Dictionary
• Afrikaans Language Resources
• Festus G2P runtime and Afrikaans models

To use this Docker image, you need to have a working version of Docker; on Windows and Mac OS you will also need a way to run a Linux kernel in a virtual machine. Visit docs.docker.com and follow the "Get Started" instructions for your platform.

Minimally you want to be able to run the official Ubuntu image, like this:

docker run --rm -it ubuntu:xenial /bin/bash

Once that is working you can try out synthesizing with Festival's KAL diphone voice:

echo Hello world |
docker run --rm -i mjansche/tts-tutorial-sltu2016 festival/bin/text2wave \
  > hello.wav

On many versions of GNU/Linux, the following also works:

echo Hello world |
docker run --rm -i mjansche/tts-tutorial-sltu2016 festival/bin/text2wave |
aplay

The software shipped with this image can also be used to build a new voice from scratch. This will be covered during the tutorial at SLTU.