AVS uses machine learning algorithms and large databases of spoken language to generate synthetic voices that can mimic human speech.