StrawNet: Self-Training WaveNet for TTS in Low-Data Regimes
The audio samples on this page go together with "StrawNet: Self-Training WaveNet for TTS in Low-Data Regimes", by Manish Sharma, Tom Kenter and Rob Clark, published in INTERSPEECH 2020.
They correspond to the single speaker experiments (section 5.1 in the paper).
They correspond to the single speaker experiments (section 5.1 in the paper).
| Sample 1: As a golfer, it’s always smart to wear 2 pairs of pants. You know, in case you get a hole in one. | |||
| #Recordings for training | WaveNet teacher | WaveNet student | StrawNet student Synthesis step involves 40k utterances |
|---|---|---|---|
| 3000 | |||
| 5000 | |||
| 7000 | |||
| 10000 | |||
| 15000 | |||
| 24000 | |||
| Sample 2: Playing Teenage Mutant Ninja Turtles dance show with April O'Neil at Nick Hotel on YouTube. | |||
| #Recordings for training | WaveNet teacher | WaveNet student | StrawNet student Synthesis step involves 40k utterances |
|---|---|---|---|
| 3000 | |||
| 5000 | |||
| 7000 | |||
| 10000 | |||
| 15000 | |||
| 24000 | |||
| Sample 3: According to Wikipedia: C. Rajagopalachari became the only Indian Governor-General. | |||
| #Recordings for training | WaveNet teacher | WaveNet student | StrawNet student Synthesis step involves 40k utterances |
|---|---|---|---|
| 3000 | |||
| 5000 | |||
| 7000 | |||
| 10000 | |||
| 15000 | |||
| 24000 | |||