Write-a-Speaker

The emotional Motion Capture dataset has been introduced in our AAAI2021 paper of “Write-a-speaker: Text-based Emotional and Rhythmic Talking-head Generation”. The demo video shows the generated talking face results relying on the dataset. Some examples of facial expression blendshapes are illustrated as below. The full 51 facial expression blendshapes can be observed in ./images.

Downloading

Please download and read the LICENSE AGREEMENT carefully, and send the license agreement with your handwritten signature to [email protected] via email. The dataset is released with Google Drive due to its oversize, so you also have to provide your Gmail in the email. Note: the permission for application is only open to researchers, faculty, engineers from universities, research institutes, and also enterprises. It is prohibited for students to apply.