search

yanggeng1995 / vae_tacotron

MIT License
51 stars 19 forks source link

Pretrained model #4

Open entn-at opened 5 years ago

entn-at commented 5 years ago

Hi Geng Yang,

thank you for providing such a nice implementation! Would you be able to upload a pretrained model (based on LJSpeech data)?

Thanks, Ewald

ZohaibAhmed commented 5 years ago

I have a model trained up to 88k: https://drive.google.com/file/d/1IkqKxT9XWWATVdyTOkJR-0AsMep4oyR7/view?usp=sharing

This is before this change though: https://github.com/yanggeng1995/vae_tacotron/commit/7d7bc3a43cdd48734f8da11e56d8ccf8e4661cf2

@yanggeng1995 - does the style transfer actually work from the reference audio? I added the reference_output change that you had here (https://github.com/yanggeng1995/vae_tacotron/commit/b0288f1caa776a98195dd94d1e8ea7ca6ec05f57#diff-f2872331652191ab856c54118796f2f4R35) in the model above, but that doesn't seem to do anything no matter what reference audio I provide.

entn-at commented 5 years ago

Thanks, @ZohaibAhmed!!