Waveform Synthesis on LJSpeech
17.02Training Time (h)FastSpeech 2
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| FastSpeech 2Vocoder=Parallel WaveGAN, Batch Size=1, Training Pipeline=Direct Training2020.06 | 17.02 | 0.0195 | 47.8 | |
| Transformer TTSVocoder=Parallel WaveGAN, Batch Size=12020.06 | 38.64 | 0.932 | — | |
| FastSpeechVocoder=Parallel WaveGAN, Batch Size=1, Training Pipeline=Teacher-Student Distillation2020.06 | 53.12 | 0.0192 | 48.5 | |
| FastSpeech 2sVocoder=End-to-End, Batch Size=12020.06 | 92.18 | 0.018 | 51.8 |