This paper describes Tacotron 2, a neural network architecture for speech synthesis directly from text. The system is composed of a recurrent sequence-to-sequence feature prediction network that maps character embeddings to mel-scale spectrograms, followed by a modified WaveNet model acting as a vocoder to synthesize timedomain waveforms from those spectrograms. Our model achieves a mean opinion score (MOS) of 4.53 comparable to a MOS of 4.58 for professionally recorded speech. To validate our design choices, we present ablation studies of key components of our system and evaluate the impact of using mel spectrograms as the input to WaveNet instead of linguistic, duration, and F_0 features. We further demonstrate that using a compact acoustic intermediate representation enables significant simplification of the WaveNet architecture.
pip3 install -r requirements.txt
1.Download and extract the LJ Speech dataset in the current directory;
First, create a directory to save output and logs.
$ mkdir outdir logdir
$ python3 train.py --output_directory=outdir --log_directory=logdir --target_val_loss=0.5
$ python3 -m multiproc train.py --output_directory=outdir --log_directory=logdir --hparams=distributed_run=True --target_val_loss=0.5
$ python3 -m multiproc train.py --output_directory=outdir --log_directory=logdir --hparams=distributed_run=True,fp16_run=True --target_val_loss=0.5
GPUs | FP16 | FPS | Score(MOS) |
---|---|---|---|
1x8 | True | 9.2 | 4.460 |
Convergence criteria | Configuration (x denotes number of GPUs) | Performance | Accuracy | Power(W) | Scalability | Memory utilization(G) | Stability |
---|---|---|---|---|---|---|---|
score(MOS):4.460 | SDK V2.2,bs:128,8x,AMP | 77 | 4.46 | 128*8 | 0.96 | 18.4*8 | 1 |
此处可能存在不合适展示的内容,页面不予展示。您可通过相关编辑功能自查并修改。
如您确认内容无涉及 不当用语 / 纯广告导流 / 暴力 / 低俗色情 / 侵权 / 盗版 / 虚假 / 无价值内容或违法国家有关法律法规的内容,可点击提交进行申诉,我们将尽快为您处理。