This is fun. The audio isn’t particularly believable. Not that it matters for this project.

I’ve used a few tts providers and although a technical marvel they are all still easily identifiable as ai. Humans are very good at spotting fakes.

This project might work better with mostly canned audio clips of real humans.