search

daniilrobnikov / vits2

VITS2: Improving Quality and Efficiency of Single-Stage Text-to-Speech with Adversarial Learning and Architecture Design
https://vits-2.github.io/demo/
MIT License
503 stars 53 forks source link

Multi-speaker training #15

Open pengpengtao opened 2 months ago

pengpengtao commented 2 months ago

When I train with data from two custom speakers, why is the resulting audio unclear, and why is there an electric noise?