Non-native English speakers read the Grandfather Passage. With explicit consent, we cloned each recording with three TTS systems and had US-English annotators rate paired source/clone clips on a 1–5 Likert scale across 7 dimensions — unaware of which clip was human or synthetically generated.
