Voicebuilding for Text-to-speech Synthesis (SS 2018)

Description

This seminar will allow the participants to build their own voices for TTS synthesis. This includes designing and recording (in a professional studio!) small speech databases and processing them with open-source tools.

We will be using MaryTTS; furthermore, familiarity with command-line shell interaction are mandatory. Bringing a laptop running MacOS X or Linux is an advantage.

Prerequisites

To be eligible for this course, participants must have successfully taken the lecture Text-to-Speech Synthesis. This means passing the final exam.

Participants must subscribe to the course mailing list, which will serve as the primary means of communication.

Slides

Schedule

This seminar will run as a block for two weeks in July/August 2018 (30.07.–10.08.2018) and require full-time participation during that period. Each day will have a morning session, with group work in the afternoons.

The course will take place at the Computational Linguistics Department (building C7.3, seminar room 1.12), with the exception of the recordings, which will be made in the studio of building C7.4.

Week 1 (30.07.–03.08.)

Time Monday Tuesday Wednesday Thursday Friday
10–13 Lecture Recording / Processing Recording / Processing Recording / Processing Recording / Processing
14–17 Lecture Recording / Processing Recording / Processing Recording / Processing Recording / Processing

Week 2 (06.08.–10.08.)

Time Monday Tuesday Wednesday Thursday Friday
10–13 Lecture Lecture Lecture Lecture Lecture
14–17 Group work Group work Group work Group work Group work