Despite impressive progress, zero-shot Text-to-Speech (TTS) models still struggle with challenging linguistic scenarios — think tongue twisters, repeated words, code-switching, and cross-lingual synthesis.