Open-weight Audio2Face-3D and Audio2Emotion networks and a sample dataset for training and evaluation