## Version 2

This release was exported from the checkpoint after 950 completed training epochs.

Compared with V1, this version has noticeably clearer sound, stronger prosody,
more natural emphasis and greater emotional variation.