search

keithito / tacotron

A TensorFlow implementation of Google's Tacotron speech synthesis with pre-trained model (unofficial)
MIT License
2.95k stars 959 forks source link

Speaker ID embedding size... #287

Open samuel-lunii opened 5 years ago

samuel-lunii commented 5 years ago

Hi @keithito ! In the multispeaker branch, I wonder why the speaker_ids embedding size is set to 377 ? Is this the result of a particular tuning or is it related to the number of speakers in the corpus ? Thanks !

Charlottecuc commented 5 years ago

Hi @samuel-lunii ! Did you figure out why the size is 377?

samuel-lunii commented 5 years ago

@Charlottecuc Did not have time to investigate yet !...

Charlottecuc commented 5 years ago

@samuel-lunii I checked the code and I think it should be the number of speakers.