Does a Cloned Voice Stay the Same? Scoring Qwen3-TTS With ECAPA

I scored every sentence chunk of two Qwen3-TTS 1.7B narrations against the voice sample with ECAPA embeddings, and re-read any chunk under 0.7.