DEMONSTRATION - SPEAKER INTERPOLATION

Our HMM-based speech synthesis system generates synthesized voices from Acoustic features of the speech data.
Therefore it can generates synthesized voices that are a mix of two or more voice characteristics.

• Interpolated voice
  "Arayuru genjitsu o subete jibun no hoo e nejimage ta no da."
  • Male 100% Female 000%
    Synthesis speech
  • Male 075% Female 025%
    Synthesis speech
  • Male 050% Female 050%
    Synthesis speech
  • Male 025% Female 075%
    Synthesis speech
  • Male 000% Female 100%
    Synthesis speech

• Morphing voice
  "osyokujiken subete yonjyuuni san sai kara onnjyuuroku shichi sai made kachouhosa kacho bijityou to itta atari ga machigai wo okosu."
  • Male -> Female
    Synthesis speech
  • Female -> Male
    Synthesis speech

Did you hear the voice that each speaker's characters was mixed?
We can create various kinds of voice character by changing various parameters.