LLM is capable of role-playing, character.ai for example can get into the role of any character after being trained. The sound is just text-to-speech, character.ai already includes that, though if a realistic voice is desired, it would need to be generated by a more sophisticated method, which is already being done. Example: Neuro-sama, ElevenLabs
Practice what you preach. Read the thread again, what do you think “say something about you” mean?