Go to file

p92587461 8fd9dd1a2b Update README.md		2023-01-29 18:33:25 +08:00
.gitmodules	ADD file via upload	2023-01-29 18:13:09 +08:00
Dockerfile	ADD file via upload	2023-01-29 18:22:24 +08:00
LICENSE	ADD file via upload	2023-01-29 18:21:37 +08:00
README.md	Update README.md	2023-01-29 18:33:25 +08:00
Tacotron2+HifiGANPaimonVocie.ipynb	ADD file via upload	2023-01-29 18:18:23 +08:00
audio_processing.py	ADD file via upload	2023-01-29 18:13:55 +08:00
data_utils.py	ADD file via upload	2023-01-29 18:22:42 +08:00
distributed.py	ADD file via upload	2023-01-29 18:22:33 +08:00
hparams.py	ADD file via upload	2023-01-29 18:22:12 +08:00
inference.ipynb	ADD file via upload	2023-01-29 18:22:01 +08:00
layers.py	ADD file via upload	2023-01-29 18:21:47 +08:00
logger.py	ADD file via upload	2023-01-29 18:21:26 +08:00
loss_function.py	ADD file via upload	2023-01-29 18:21:16 +08:00
loss_scaler.py	ADD file via upload	2023-01-29 18:21:05 +08:00
model.py	ADD file via upload	2023-01-29 18:20:45 +08:00
multiproc.py	ADD file via upload	2023-01-29 18:20:19 +08:00
plotting_utils.py	ADD file via upload	2023-01-29 18:20:05 +08:00
requirements.txt	ADD file via upload	2023-01-29 18:19:32 +08:00
stft.py	ADD file via upload	2023-01-29 18:19:14 +08:00
train.py	ADD file via upload	2023-01-29 18:18:43 +08:00
utils.py	ADD file via upload	2023-01-29 18:18:53 +08:00

PaimonVoice

来源：北京理工大学CSTAIG研究组人工智能练习

我们制作了一个基于tacotron2和HiFiGAN声码器的语音合成模型，在google colab上进行模型的训练和测试。我们语音合成对象是游戏《原神》中的角色派蒙代码部分上传不完全，所有工程文件和资源如下：

Tacotron2+HiFiGAN打包链接：https://pan.baidu.com/s/1ngCUvifQM6ETwuG-NFQeGA 提取码：z6h3

400条派蒙语音测试集链接：https://pan.baidu.com/s/1g0C0Ck4P_BxdTgMirKRc-g 提取码：5ew1

1800条派蒙语音训练集链接：https://pan.baidu.com/s/1IDA4lppAJGHnophQAwkxNg 提取码：f2xk

将所有文件夹放到Google云盘MyDrive下，然后通过colab打开位于MyDrive/tacotron2/Tacotron2+HifiGAN派蒙语音合成.ipynb

如果需要训练新的模型，依次运行代码即可。
如果只是利用现有的模型（存放于MyDrive/colab/outdir）合成语音，只需Data Preparation、Download Tacotron 2和最后的With HiFi-GAN即可