Humor is a defining feature of musicals, functioning as a key device to relieve tension, shape character identity, and sustain audience engagement. Translating humor for musical subtitles is particularly demanding since translations must fit strict spatial and temporal constraints while also synchronizing with multimodal performance cues such as gesture, rhythm, and stage design. This study compares Korean stage subtitles of the licensed musical 9 to 5 produced by ChatGPT (GPT-5) and by professional human translators. A total of 76 humorous instances were collected and categorized into language-dependent, culture-dependent, character-dependent, and sexual humor. Findings reveal clear contrasts. AI translation consistently respected character limits and produced fluent literal renderings, yet failed to reproduce phonetic play (rhyme, alliteration), to creatively adapt culture-specific jokes (units, currency, brand names), and to maintain pragmatic consistency in tone and relationships. It also relied heavily on exclamations while neglecting suffixes, slang, or neologisms, and overlooked multimodal cues such as gestures and visual symbols. Human subtitles, by contrast, drew on explicitation, compensation, and creative substitution, reinforcing comic effect with icons, fonts, and onomatopoeia in sync with stage action. Beyond identifying limitations, this study proposes specific strategies for human translators when working with AI outputs: pre-specifying rhyme or pun candidates, restoring elided elements, and standardizing character voices. In particular, translators should intervene more actively in cultural and sexual humor, adapting references and softening taboos to match audience reception. Overall, the findings suggest a complementary workflow in which AI supports draft generation and constraint management, while human creativity ensures cultural resonance, multimodal coherence, and effective humor delivery.
1. 들어가며
2. 이론적 배경
3. 연구 방법
4. 결과
5. 나가며
인용문헌
(0)
(0)