Sound2Sight: Generating Visual Dynamics from Sound and Context

TL;DR

Sound2Sight introduces a deep variational framework that generates future video frames conditioned on audio and past frames, enabling diverse and coherent visual predictions in multimodal reasoning tasks.

Contribution

It presents a novel stochastic prior conditioned on audio-visual embeddings and a multimodal discriminator, improving video synthesis quality and diversity.

Findings

Outperforms prior methods in video quality and diversity

Effective in occlusion reasoning scenarios

Validated on multiple datasets including new ones

Abstract

Learning associations across modalities is critical for robust multimodal reasoning, especially when a modality may be missing during inference. In this paper, we study this problem in the context of audio-conditioned visual synthesis -- a task that is important, for example, in occlusion reasoning. Specifically, our goal is to generate future video frames and their motion dynamics conditioned on audio and a few past frames. To tackle this problem, we present Sound2Sight, a deep variational framework, that is trained to learn a per frame stochastic prior conditioned on a joint embedding of audio and past frames. This embedding is learned via a multi-head attention-based audio-visual transformer encoder. The learned prior is then sampled to further condition a video forecasting module to generate future frames. The stochastic prior allows the model to sample multiple plausible futures that are consistent with the provided audio and the past context. Moreover, to improve the quality and coherence of the generated frames, we propose a multimodal discriminator that differentiates between a synthesized and a real audio-visual clip. We empirically evaluate our approach, vis-\'a-vis closely-related prior methods, on two new datasets viz. (i) Multimodal Stochastic Moving MNIST with a Surprise Obstacle, (ii) Youtube Paintings; as well as on the existing Audio-Set Drums dataset. Our extensive experiments demonstrate that Sound2Sight significantly outperforms the state of the art in the generated video quality, while also producing diverse video content.