home page
home page

Back to menu

Euromasters summer school 2005

Tutorial 3: Unit selection speech synthesis using Festival

Rob Clark, Korin Richmond and Simon King, University of Edinburgh, UK

Festival is a complete toolkit for TTS and is the most widely used system in speech synthesis research around the world. This is a hands on tutorial that covers the basics of using Festival and the background required to develop new voices for Festival.

Students will build a fully functional open-domain unit-selection voice for Festival 2.0 using the brand new Multisyn voice building tools and unit selection engine. We will provide the recorded speech and students will see all the steps involved it producing a "voice" from this raw data.

The techniques used in Festival 2.0 are similar to those in the major commercial systems, like those from Rhetorical Systems or AT&T.

With the voice building tools included in Festival 2.0, very high quality voices can be built relatively quickly, without tedious manual annotation of the data.

Here are some examples of the quality you can acheive with Festival 2.0

Back to registration page

Location of tutorial files

/group/cstr/projects/euromasters/tutorial3 on the Informatics linux machines ("DICE" machines)