Tacotron + WaveRNN synthesis
Makes use of:
- Tacotron: https://github.com/Rayhane-mamah/Tacotron-2
- WaveRNN: https://github.com/fatchord/WaveRNN
You'll at least need python3, PyTorch 0.4.1, Tensorflow and librosa.
python3 preprocess.py --model='WaveRNN'
Default parameters:
| name | default | |
|---|---|---|
| --base_dir | ||
| --hparams | ex) 'wavernn_gpu_num=4, wavernn_batch_size=16' | |
| --model | 'Tacotron' | 'Tacotron', 'WaveRNN' |
| --dataset | 'LJSpeech-1.1' | 'LJSpeech-1.0', 'LJSpeech-1.1', 'M-AILABS' |
Others, look at this file...
python3 train.py --model='Tacotron-2' --GTA --use_cuda
If you would like to train separately...
# Tacotron
python3 train.py --model='Tacotron'
# Tacotron synth
python3 synthesize.py --model='Tacotron' --mode='synthesis' --GTA
# WaveRNN
python3 train.py --model='WaveRNN' --use_cuda
Default parameters:
| name | default | |
|---|---|---|
| --base_dir | ||
| --hparams | ex) 'wavernn_gpu_num=4, wavernn_batch_size=16' | |
| --model | 'Tacotron-2' | 'Tacotron-2', 'Tacotron', 'WaveRNN' |
| --mode | 'synthesis' | 'eval', 'synthesis', 'live' |
| --init | False | True, False |
| --slack_url | {your slack wabhook url...} | |
| --use_cuda | False | True, False |
Others, look at this file...
python3 synthesize.py --model='Tacotron-2' --text_list={your text file}
Default parameters:
| name | default | |
|---|---|---|
| --base_dir | ||
| --hparams | ex) 'wavernn_gpu_num=4, wavernn_batch_size=16' | |
| --model | 'Tacotron-2' | 'Tacotron-2', 'Tacotron', 'WaveRNN' |
| --mode | 'eval' | 'eval', 'synthesis', 'live' |
| --text_list | {your text file...} | |
| --use_cuda | False | True, False |
Others, look at this file...
https://github.com/h-meru/Tacotron-WaveRNN/files/2444777/wavernn_model.zip
https://github.com/h-meru/Tacotron-WaveRNN/files/2444792/Samples_730k.zip
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent