pytorch_voice_conversion

use pytorch to implement voice conversion

ToDo List

• GMM Baseline
• LSTM Result
• Dual Result
• Gan Result

First

check the data root in config.py and replace it by yours

GMM Baseline

the gmm baseline is copied from the nnmnkwii gitpage

LSTM

Prepare Data

replace the in_dir and out_dir in run_pre.sh and then run it.

the in_dir contains wavs like xx/arctic/cmu_us_bdl_arctic/wav/xx.wav

Train

replace the data_root in run_rnn.sh and choose the source(target) speaker, then run it.

finally the checkpoint will be saved in the dir checkpoints

Test

replace the ssp tsp checkpoint_path in test.py, the result wav will be under the dir wavs

Dual

the train and test process is the same of LSTM except add the args dual