This page contains following samples:
- Natural: Originally recorded voice.
- Vocoded: Use STRAIGHT to extract 60-dimensional Mel-Cepstral Coefficients (MCCs), 25 band aperiodicities (BAPs) and fundamental frequency (F0), and then pass these parameters directly to STRAIGHT-vocoder to reconstract waveform.
- Uni-Grapheme: Synthesized voice from informal transliteration using graphemes.
- Multi-Grapheme: Synthesized voice from informal transliteration using multi-grapheme model.
- G2P: Synthesized voice from informal transliteration using G2P model.
- Benchmark: Synthesized voice from automated transliterated text(from native text) using common phone set notation.