Large-scale unsupervised audio pre-training for video-to-speech synthesis

TL;DR

This paper introduces a large-scale unsupervised audio pre-training approach for video-to-speech synthesis, leveraging extensive audio-only datasets to improve speech reconstruction quality in a cross-modal setting.

Contribution

It proposes pre-training audio decoders on large audio-only datasets to enhance video-to-speech synthesis, enabling the use of abundant audio data without requiring visual annotations.

Findings

Pre-training improves speech reconstruction quality.

Unsupervised audio pre-training benefits cross-modal synthesis.

Method outperforms existing approaches on benchmarks.

Abstract

Video-to-speech synthesis is the task of reconstructing the speech signal from a silent video of a speaker. Most established approaches to date involve a two-step process, whereby an intermediate representation from the video, such as a spectrogram, is extracted first and then passed to a vocoder to produce the raw audio. Some recent work has focused on end-to-end synthesis, whereby the generation of raw audio and any intermediate representations is performed jointly. All such approaches involve training on data from almost exclusively audio-visual datasets, i.e. every audio sample has a corresponding video sample. This precludes the use of abundant audio-only datasets which may not have a corresponding visual modality (e.g. audiobooks, radio podcasts, speech recognition datasets etc.), as well as audio-only architectures that have been developed by the audio machine learning community over the years. In this paper we propose to train encoder-decoder models on more than 3,500 hours of audio data at 24kHz, and then use the pre-trained decoders to initialize the audio decoders for the video-to-speech synthesis task. The pre-training step uses audio samples only and does not require labels or corresponding samples from other modalities (visual, text). We demonstrate that this pre-training step improves the reconstructed speech and that it is an unexplored way to improve the quality of the generator in a cross-modal task while only requiring samples from one of the modalities. We conduct experiments using both raw audio and mel spectrograms as target outputs and benchmark our models with existing work.