AI 생성 더빙 번역 특성 연구 : 중국 신화 숏폼 드라마 《Ascension to the Moon(항아분월)》을 중심으로

다국어 초록 (Multilingual Abstract)

In the context of an ever-more globalized society, the cross-border dissemination of traditional Chinese culture has become a strategic priority. Dubbing translation, a crucial link in audiovisual translation （AVT）, directly affects overseas audiences’ understanding of and connection with Chinese cultural content. Advancements in AI-based translation have opened up new possibilities for dubbing-related language processing. However, existing research has largely focused on technical algorithm optimization, with limited empirical analysis of traditional cultural content, especially regarding emotional delivery, cultural localization, and prefabricated orality.
This study systematically examines the characteristics of AI-generated English dubbing scripts through a case study ofAscension to the Moon, China’s first fully AI-produced （art, storyboarding, dubbing, and scoring） microdrama, which depicts a classic chapter fromChinese Mythology. AI-generated English dubbing texts serve as the primary subject for analysis. Grounded in dubbing theory, AVT, and multimodality, the analysis focuses on three dimensions: emotional delivery, cultural localization, and prefabricated orality. Six representative utterances from key narrative segments are qualitatively analyzed based on these dimensions.
The findings indicate that, in this case, AI-generated English dubbing scripts exhibit relatively consistent tendencies. In emotional delivery, they build layered structures through specific lexical choices （e.g., “gazes” vs. “looks”）, scene contrasts （e.g., earthly reunion vs. lunar solitude）, and syntactic rhythms, though narrative pacing tends to be slower and mixed emotions remain challenging. In cultural localization, AI preserves the core Chinese cultural connotations while adapting them for English audiences using explanatory strategies （e.g., “elixir of immortality” for xiandan, “a day of reunion” for tuanyuan）. Core symbols are consistently translated, but cultural metaphors may be diluted, limiting accessibility. In terms of prefabricated orality, AI output generally follows English grammatical rules and idiomatic expressions, employing complex sentences and high-frequency collocations （e.g., “homes light up”, “passed down through generations”）. However, sentence structures tend to be less flexible, lacking ellipsis, inversion, anddiscourse connectives typical of spontaneous speech, resulting in a relatively formal style.
The study suggests that AI-generated English dubbing scripts may be suitable for cultural content aimed at audiences who already possess a certain level of cultural comprehension （e.g., academic presentations, cultural exchanges）. For fast-paced, entertainment-oriented mass communication, improvements in conciseness and colloquial naturalness may be needed.
This research serves as empirical evidence for the understudied area of AI-generated dubbing script translation for works rooted in traditional culture. Limitations include the small sample size of a single case and a single language pair （Chinese-English）, a lack of data on audience reception, and restricted access to AI system details. Future research should expand to diverse text types, multiple language pairs, empirical audience studies, and a deeper analysis of the technical process involved.

목차 (Table of Contents)

• 표 목차 iii
• 그림 목차 iv
• I. 서론 1
• 1.1. 연구 배경 및 목적 1
• 1.2. 연구 중요성 4
• 1.3. 연구 대상 및 분석 방법 8
• II. 이론적 배경 11
• 2.1. 더빙 정의 11
• 2.2. 시청각번역（AVT）과 멀티모달 관점 12
• 2.3. AI 더빙 대사 번역의 핵심 요소에 대한 연구 14
• 2.3.1. 감정 전달 14
• 2.3.2. 문화 현지화 15
• 2.3.3. 가공된 구어성 18
• 2.4. AI 생성 영어 더빙 대사 번역의 품질 평가 연구 19
• III. 연구방법 21
• 3.1. 연구 사례 분석 21
• 3.1.1. 감정 전달 22
• 3.1.2. 문화 현지화 25
• 3.1.3. 가공된 구어성 28
• 3.2. 연구 분석 결과 31
• IV. 결론 36
• 4.1. 연구결과 요약 36
• 4.2. 연구의 의의 39
• 4.3. 연구의 한계 및 향후 과제 43
• 참고문헌 48
• 부록 1. 추가 예시 분석 설명（감정 전달） 51
• 부록 2. 추가 예시 분석 설명（문화 현지화） 53
• 부록 3. 추가 예시 분석 설명（가공된 구어성） 55
• ABSTRACT 57