Audio samples from "Neural speech-rate conversion with multispeaker WaveNet vocoder"

Authors: T. Okamoto, K. Matsubara, T. Toda, Y. Shida and H. Kawai

Experiment 1 (Japanese female: F117)

Experiment 2 (Japanese female: JVS004)

Experiment 2 (Japanese male: JVS001)

Experiment 3 (English female: slt)

Experiment 3 (English male: bdl)

Experiment 4 (Japanese female: F009)

Experiment 1 (Japanese female: F117)

Normal rateFast rate (x 0.8)Slow rate (x 1.5)
Original
WSOLA
STRAIGHT
Single-speaker WaveNet
Single-speaker WaveNet (a)
Single-speaker WaveNet (b)
Multispeaker WaveNet
Multispeaker WaveNet (a)
Multispeaker WaveNet (b)

Experiment 2 (Japanese female: JVS004)

Normal rateFast rate (x 0.8)Slow rate (x 1.5)
Original
WSOLA
STRAIGHT
Multispeaker WaveNet
Multispeaker WaveNet (b)

Experiment 2 (Japanese male: JVS001)

Normal rateFast rate (x 0.8)Slow rate (x 1.5)
Original
WSOLA
STRAIGHT
Multispeaker WaveNet
Multispeaker WaveNet (b)

Experiment 3 (English female: slt)

Normal rateFast rate (x 0.8)Slow rate (x 1.5)
Original
WSOLA
STRAIGHT
Single-speaker WaveNet
Single-speaker WaveNet (a)
Multispeaker WaveNet
Multispeaker WaveNet (b)

Experiment 3 (English male: bdl)

Normal rateFast rate (x 0.8)Slow rate (x 1.5)
Original
WSOLA
STRAIGHT
Single-speaker WaveNet
Single-speaker WaveNet (a)
Multispeaker WaveNet
Multispeaker WaveNet (b)

Experiment 4 (Japanese female: F009)

Normal rateFast rateSlow rate
Original
WSOLA
STRAIGHT (Oracle acoustic feature)
STRAIGHT (Resampled acoustic feature)
Multispeaker WaveNet (Oracle acoustic feature)
Multispeaker WaveNet (b)