Self-Supervised Audio-Visual Soundscape Stylization

TL;DR

This paper introduces a self-supervised model that manipulates speech sounds to match the acoustic environment of a given scene using unlabeled audio-visual data, enhancing sound transfer and stylization capabilities.

Contribution

It presents a novel self-supervised approach for audio-visual soundscape stylization using latent diffusion models trained on unlabeled videos.

Findings

Model successfully transfers sound properties between scenes.

Visual signals improve sound prediction accuracy.

Effective training with unlabeled in-the-wild videos.

Abstract

Speech sounds convey a great deal of information about the scenes, resulting in a variety of effects ranging from reverberation to additional ambient sounds. In this paper, we manipulate input speech to sound as though it was recorded within a different scene, given an audio-visual conditional example recorded from that scene. Our model learns through self-supervision, taking advantage of the fact that natural video contains recurring sound events and textures. We extract an audio clip from a video and apply speech enhancement. We then train a latent diffusion model to recover the original speech, using another audio-visual clip taken from elsewhere in the video as a conditional hint. Through this process, the model learns to transfer the conditional example's sound properties to the input speech. We show that our model can be successfully trained using unlabeled, in-the-wild videos, and that an additional visual signal can improve its sound prediction abilities. Please see our project webpage for video results: https://tinglok.netlify.app/files/avsoundscape/