Triple M: A Practical Text-to-speech Synthesis System With Multi-guidance Attention And Multi-band Multi-time LPCNet

TL;DR

This paper introduces Triple M, a TTS system combining multi-guidance attention for improved naturalness and a multi-band multi-time vocoder for enhanced efficiency, suitable for large-scale online deployment.

Contribution

The paper presents a novel multi-guidance attention mechanism and an efficient multi-band multi-time LPCNet vocoder, advancing TTS quality and computational efficiency.

Findings

26.8% reduction in word error rate

Speeds up LPCNet by 2.75 times on CPU

Reduces computational complexity from 2.8 to 1.0 GFLOP

Abstract

In this work, a robust and efficient text-to-speech (TTS) synthesis system named Triple M is proposed for large-scale online application. The key components of Triple M are: 1) A sequence-to-sequence model adopts a novel multi-guidance attention to transfer complementary advantages from guiding attention mechanisms to the basic attention mechanism without in-domain performance loss and online service modification. Compared with single attention mechanism, multi-guidance attention not only brings better naturalness to long sentence synthesis, but also reduces the word error rate by 26.8%. 2) A new efficient multi-band multi-time vocoder framework, which reduces the computational complexity from 2.8 to 1.0 GFLOP and speeds up LPCNet by 2.75x on a single CPU.