LJ Speech & ESD model

Release contents:

• Daft-Exprt model pre-trained on LJ Speech Dataset and the Emotional Speech Dataset from Zhou et al.
• Prosody transfer examples synthesized using this pre-trained model and Griffin-Lim algorithm

Full disclosure: The model provided in this release is not the same as in the paper evaluation. The model of the paper was trained with proprietary data which prevents us to release it publicly.