List your company
ResearchFormerly WizWand

Speech Synthesis on LJ Speech (test)

4.54MOSGT
2.07522.71513.3553.9949Dec 3, 2019Jul 18, 2020Mar 4, 2021Oct 19, 2021Jun 5, 2022Jan 20, 2023Sep 6, 2023
Updated 8mo ago

Evaluation results

MethodLinks
GTDescription=Ground Truth2020.054.54——
GT2022.044.52——
Gaussian WaveNetflows x layers=1x30 = 30, res. channels=128, # param=4.57 M, syn. speed=0.002x2019.124.43——
WaveFlowh=16, flows x layers=8x8 = 64, res. channels=256, # param=86.18 M, syn. speed=8.42x2019.124.43——
WaveFlowh=16, flows x layers=8x8=64, res. channels=128, # param=22.25 M, syn. speed=21.32x2019.124.38——
WaveGlowflows x layers=12x8=96, res. channels=256, # param=87.88 M, syn. speed=34.69x2019.124.34——
WaveFlowh=16, flows x layers=8x8=64, res. channels=96, # param=12.78 M, syn. speed=26.23x2019.124.34——
WaveGlowflows x layers=12x8=96, res. channels=512, # param=268.29 M, syn. speed=8.08x2019.124.32——
WaveFlowh=16, flows x layers=8x8=64, res. channels=64, # param=5.91 M, syn. speed=42.60x2019.124.32——
GT(voc.)Vocoder=FastDiff2022.044.28——
WaveFlowh=8, flows x layers=8x8=64, res. channels=64, # param=5.91 M, syn. speed=47.61x2019.124.26——
ClariNetflows x layers=6x10=60, res. channels=64, # param=2.17 M, syn. speed=21.64x2019.124.22——
GTDescription=Ground Truth Mel-spectrogram, Vocoder=WaveGlow2020.054.19——
CascadedAcoustic Model=PortaSpeech, Vocoder=FastDiff2022.044.13——
VOCParams=13.9M2023.094.130.0011.97
FastDiff-TTS2022.044.03——
Glow-TTSTemperature (T)=0.333, Vocoder=WaveGlow2020.054.01——
Glow-TTSTemperature (T)=0.667, Vocoder=WaveGlow2020.053.97——
Glow-TTSTemperature (T)=0.500, Vocoder=WaveGlow2020.053.96——
FastSpeech 2s2022.043.94——
Tacotron 2Vocoder=WaveGlow2020.053.88——
Matcha-TTSParams=18.2M, RAM=4.8 GiB, NFE=102023.093.840.0382.09
Matcha-TTSParams=18.2M, RAM=4.8 GiB, NFE=42023.093.770.0192.15
VITSParams=36.3M, RAM=12.4 GiB2023.093.710.0742.52
WaveGrad 22022.043.68——
Matcha-TTSParams=18.2M, RAM=4.8 GiB, NFE=22023.093.650.0152.34
GCFMParams=14.8M, RAM=7.8 GiB, NFE=42023.093.570.0192.7
GRAD-TTSParams=14.8M, RAM=7.8 GiB, NFE=102023.093.490.0493.44
Tacotron 2Condition=T2+P, Size=28.2M2021.083.41——
FS2Params=41.2M, RAM=6.0 GiB2023.093.290.014.18
Tacotron 2Condition=T2-P, Size=23.8M2021.083.25——
Neural HMMCondition=NH2, Size=15.3M2021.083.24——
GRAD-TTSParams=14.8M, RAM=7.8 GiB, NFE=42023.093.20.0193.69
WaveGlowflows x layers=12x8 = 96, res. channels=128, # param=34.83 M, syn. speed=69.88x2019.122.97——
Neural HMMCondition=NH1, Size=12.7M2021.082.68——
WaveGlowflows x layers=12x8 = 96, res. channels=64, # param=17.59 M, syn. speed=93.53x2019.122.17——