The synthesis of natural emotional reactions is an essential criterion in vivid talking-face video generation. This criterion is nevertheless seldom taken into consideration in previous works due to the absence of a large-scale, high-quality emotional audio-visual dataset. To address this issue, we build the Multi-view Emotional Audio-visual Dataset (MEAD), a talking-face video corpus featuring 60 actors and actresses talking with eight different emotions at three different intensity levels. High-quality audio-visual clips are captured at seven different view angles in a strictly-controlled environment. Together with the dataset, we release an emotional talking-face generation baseline that enables the manipulation of both emotion and its intensity. Our dataset could benefit a number of different research fields including conditional generation, cross-modal understanding and expression recognition.

MEAD: A Large-scale Audio-visual Dataset for Emotional Talking-face Generation Kaisiyuan Wang*, Qianyi Wu*, Linsen Song*, Zhuoqian Yang, Wayne Wu+, Chen Qian, Ran He, Yu Qiao, Chen Change Loy European Conference on Computer Vision, ECCV 2020.

We build the Multi-view Emotional Audio-visual Dataset (MEAD), a talking-face video corpus featuring 60 actors talking with eight different emotions at three different intensity levels (except for neutral). The videos are simultaneously recorded at seven different perspectives in a strictly-controlled environment to provide high-quality details of facial expressions. About 40 hours of audio-visual clips are recorded for each person and view. Part0 consists 48 actors' data and Part1 consists the left 12 ones'. We have released Part0 and will release the Part1 on about June 2021.