use pytorch to implement voice conversion
- GMM Baseline
- LSTM Result
- Dual Result
- Gan Result
check the data root in config.py
and replace it by yours
the gmm baseline is copied from the nnmnkwii gitpage
replace the in_dir
and out_dir
in run_pre.sh
and then run it.
the
in_dir
contains wavs likexx/arctic/cmu_us_bdl_arctic/wav/xx.wav
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
replace the ssp tsp checkpoint_path in test.py
, the result wav will be under the dir wavs
the train and test process is the same of LSTM except add the args dual