Hundreds of hours of manually annotated emotionally-charged audiovisual multimodal datasets.