Facebook wav2vec is a tool that can be used to automatically recognize emotions from audio recordings. It is based on the idea that raw audio can be represented as a sequence of discrete events, and that these events can be learned using a neural network. The tool can be used to learn representations for speech, music, and other types of audio The tool can be used to recognize a variety of emotions, including happiness, sadness, anger, and neutral.
In this demo, we only consider URDU language. For good result record 5 seconds audio