Skip to main content

AdaSpeech fine-tuning

Created on June 24|Last edited on June 27

Samples

Freemans voice is not reproduced when only fine-tuning CLN layes and speaker embedding.

This set of panels contains runs from a private project, which cannot be shown in this report


Low quality of eval samples??




Fine-tuning additional layers

Additionally include to trainable parameters:
  • Variance adaptor
  • All decoder parameters

Run set
3