search

ebadawy / voice_conversion

MIT License
129 stars 36 forks source link

Generating audio after model training #1

Closed Himanshu-KF closed 3 years ago

Himanshu-KF commented 3 years ago

How to generate or convert is not mentioned in the README after model training is done. Can you please add that to readme as well?

Great work on this! I saw the results and they look promising.

ebadawy commented 3 years ago

Thanks for the suggestion. We are working on that and it should be added soon.

RussellSB commented 3 years ago

Hi @Himanshu-KF just to let you know a inference.py has been added for this! It accepts an input wav, and saves a wav in the style of the target speaker. The saved wav is a reconstruction from Griffin Lim, then you can pass the wav to vocoders like WaveNet or MelGAN to improve its quality.

If you prefer, there's also an option to specify an input directory of wav files, to infer in bulk on a test set of data.