Hume's Custom Model API integrates dynamic patterns of facial expression, vocal expression, and/or language into a custom multimodal model, leveraging insights gained by pretraining on millions of videos and audio files to predict your labels as accurately as possible, even after seeing just a few dozen new examples.