Diagnose, Then Refine: A Closed-Loop TTS System with AudioLLM-Guided Correction
Human evaluation on diagnosed low-quality TTS outputs diagnosed by an AudioLLM shows that LoopTTS can detect perceptually salient errors and correct them with the Refiner, outperforming raw generated audio and practical open-loop re-generation baselines in recovery quality.
Ze-Yang Song, Tianyu Liu, Tian-Rui Wang et al.
· 1 citation