Talker normalization enables listeners to adapt to speaker-specific acoustic variability, thereby facilitating speech perception across different talkers. While previous research suggests that talker normalization can occur in both Mandarin and Cantonese speech contexts, the extent to which it applies to non-linguistic contexts remains unclear. This study examined talker normalization in Mandarin-Cantonese bilinguals, comparing its effects in speech and nonspeech contexts. We recruited 36 bilingual participants to complete a forced-choice word identification task in one Mandarin paradigm and two Cantonese paradigms. The results revealed that under all three paradigms, tone normalization occurred robustly in the speech condition but was absent in the nonspeech condition. These findings provide novel evidence that talker normalization operates primarily within linguistic boundaries, supporting the view that talker normalization is language-specific rather than domain-general.