Synthesis corpus (TP8)

 
Goal for this project is to create a public available synthesis corpus for German. To find speakers with a fitting voice for synthesis we organized a casting in which professional speakers had to read a list of words (see 'List of prompts for the casting'). The material herefore was taken out of all diphones of three target sentences, namely 'Heute ist schönes Frühligswetter, 'Der herrische Pate versteht sich als Pol der ganzen Familie' and 'Wer muß noch Schularbeiten machen?'. For every diphone (e.g. p_a:) a logatom and a known word were created manually after special criteria and within those words the relevant diphones were embedded. Out of every of these spoken words the diphones were segmented and put together in a way that results in the three target sentences again. Further steps of this project are the automatical creation of all diphones used in German language and the logatoms in which they appear. We would like to thank all the speakers who helped us to create these first target sentences. The synthetic sentences can be heard under the point 'Synthetic sentences'.