Controllable Emphasis with zero data for text-to-speech

TL;DR

This paper introduces a scalable, annotation-free method for emphasizing words in text-to-speech synthesis by increasing predicted durations, improving naturalness and emphasis detection across multiple languages.

Contribution

The authors propose a simple duration-based emphasis technique that does not require recordings or annotations, outperforming spectrogram modification methods.

Findings

Improves naturalness by 7.3% over baseline

Increases correct emphasis detection by 40%

Effective across four languages and multiple voices

Abstract

We present a scalable method to produce high quality emphasis for text-to-speech (TTS) that does not require recordings or annotations. Many TTS models include a phoneme duration model. A simple but effective method to achieve emphasized speech consists in increasing the predicted duration of the emphasised word. We show that this is significantly better than spectrogram modification techniques improving naturalness by $7.3\%$ and correct testers' identification of the emphasized word in a sentence by $40\%$ on a reference female en-US voice. We show that this technique significantly closes the gap to methods that require explicit recordings. The method proved to be scalable and preferred in all four languages tested (English, Spanish, Italian, German), for different voices and multiple speaking styles.