More Than Words: In-the-Wild Visually-Driven Prosody for Text-to-Speech

Michael Hassid, Michelle Tadmor Ramanovich, Brendan Shillingford, Miaosen Wang, Ye Jia, Tal Remez; Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022, pp. 10587-10597

Abstract

In this paper we present VDTTS, a Visually-Driven Text-to-Speech model. Motivated by dubbing, VDTTS takes advantage of video frames as an additional input alongside text, and generates speech that matches the video signal. We demonstrate how this allows VDTTS to, unlike plain TTS models, generate speech that not only has prosodic variations like natural pauses and pitch, but is also synchronized to the input video. Experimentally, we show our model produces well-synchronized outputs, approaching the video-speech synchronization quality of the ground-truth, on several challenging benchmarks including "in-the-wild" content from VoxCeleb2. Supplementary demo videos demonstrating video-speech synchronization, robustness to speaker ID swapping, and prosody, presented at the project page.

Related Material

[pdf] [supp] [arXiv]
[bibtex]
@InProceedings{Hassid_2022_CVPR, author = {Hassid, Michael and Ramanovich, Michelle Tadmor and Shillingford, Brendan and Wang, Miaosen and Jia, Ye and Remez, Tal}, title = {More Than Words: In-the-Wild Visually-Driven Prosody for Text-to-Speech}, booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)}, month = {June}, year = {2022}, pages = {10587-10597} }