Tacotron 2

Reported on 6 benchmarks across 4 tasks · 2 papers · 3 SOTA

Note: results are matched by exact model name. Different papers may use the same name for different model variants.

Audio6 results

Speech RecognitiononNorth American English
Mean Opinion Score· 2017-12-16
4.526
SOTA
Natural TTS Synthesis by Conditioning WaveNet on Mel Spectrogram Predictions arXiv:1712.05884
Speech SynthesisonNorth American English
Mean Opinion Score· 2017-12-16
4.526
SOTA
Natural TTS Synthesis by Conditioning WaveNet on Mel Spectrogram Predictions arXiv:1712.05884
Accented Speech RecognitiononNorth American English
Mean Opinion Score· 2017-12-16
4.526
SOTA
Natural TTS Synthesis by Conditioning WaveNet on Mel Spectrogram Predictions arXiv:1712.05884
Text-To-Speech SynthesisonLJSpeech
Pleasantness MOS· uses extra data· 2020-05-12
3.521
best: 3.665 (Flowtron)
Flowtron: an Autoregressive Flow-based Generative Network for Text-to-Speech Synthesis arXiv:2005.05957
Text-To-Speech SynthesisonHUI speech corpus
Mean Opinion Score
3.74
Text-To-Speech SynthesisonThorsten voice 21.02 neutral
Mean Opinion Score
3.49