More interesting speech material

7 minute 5 seconds

Audio examples

3.1 Emotional synthesis using unit selection, by blending units from different databases Gregor Hofer, “Emotional Speech Synthesis“, MSc thesis, University of Edinburgh, 2004. Using only data in a single emotion Kirsten: neutral Kirsten: angry Kirsten: happy Korin: neutral Korin: angry Blending data in an emotion and neutral speech in varying proportions Kirsten: neutral Kirsten: full on angry Kirsten: half angry Kirsten: somewhat angry Kirsten: full on happy Kirsten: half happy Kirsten: somewhat happy 3.2 Building voices from read speech vs. spontaneous speech Building the voice only from spontaneous speech. The quality is not good. Building the voice only from read speech (the voice talent read out the transcript of the previously-recorded spontaneous speech). The quality is as expected from the standard HMM-based approach using this amount of data. 3.5 Building voices from read speech vs. spontaneous speech, or blending the two types of data Sebastian Andersson, Junichi Yamagishi, and Robert A.J. Clark. Synthesis and evaluation of conversational characteristics in HMM-based speech synthesis. Speech Communication, 54(2):175-188, 2012. DOI: 10.1016/j.specom.2011.08.001 HMM-based synthesis. Refer to paper for descriptions of each system below. Original natural spontaneous speech Read Aloud HTS Spontaneous HTS Blend.Read Blend.Spon