Facial expressions generation for virtual assistants

The current progress of virtual assistantants makes them useful in everyday life. However they interact with human mostly through text or audio. Giving them a visual form makes interaction with them more amusing. Therefore there is the need to create realistic movements in particular facial expressions consistent with speech. This paper proposes a system of automatic facial animation from audio. The proposed approach has a fast running time and does not require significant computing resources. Also, we discover some properties of the human perception of the talking head. Thus, we are able to overcome the uncanny valley effect and makes automatic animation more realistic.