Accelerating Codec-based Speech Synthesis with Multi-Token Prediction and Speculative Decoding

TL;DR

This paper introduces a novel inference method for codec-based speech synthesis that predicts multiple tokens simultaneously and employs speculative decoding, significantly speeding up synthesis with minimal quality loss.

Contribution

It proposes a multi-token prediction approach combined with speculative decoding to accelerate speech synthesis without additional training.

Findings

Prediction time reduced by 4-5 times

Minimal quality degradation or improved speech intelligibility

Flexible speed-quality trade-offs during inference

Abstract

The goal of this paper is to accelerate codec-based speech synthesis systems with minimum sacrifice to speech quality. We propose an enhanced inference method that allows for flexible trade-offs between speed and quality during inference without requiring additional training. Our core idea is to predict multiple tokens per inference step of the AR module using multiple prediction heads, resulting in a linear reduction in synthesis time as the number of heads increases. Furthermore, we introduce a novel speculative decoding technique that utilises a Viterbi-based algorithm to select the optimal sequence of generated tokens at each decoding step. In our experiments, we demonstrate that the time required to predict each token is reduced by a factor of 4 to 5 compared to baseline models, with minimal quality trade-off or even improvement in terms of speech intelligibility. Audio samples are available at: multpletokensprediction.github.io/multipletokensprediction.github.io/.