“AI voice” can describe several different things. A speech generator turns text into audio. Voice cloning attempts to reproduce a particular voice from samples. A live voice effect changes microphone audio during a conversation. The tools, quality and rights involved can differ substantially.
If a reasonable listener could think the speech came from a real person, get that person's permission before creating or publishing an imitation. Clearly label synthetic speech when it matters to the audience. Keep a record of the license or permission for voices and sound clips you use.
For live audio routing, start with real-time voice effects. For clearer original speech, start with microphone quality.