Deepseek, ChatGPT, and Gemini versus human subtitling: a case study in socio-cultural adaptation in multimedia communication

Effective subtitling of culturally sensitive content requires deep linguistic and socio-cultural expertise. This study examines the translation of religious, sociocultural, and political elements in a television series from English to Persian, comparing a human subtitler's work with three generative AI models: DeepSeek, ChatGPT, and Gemini. Using a qualitative approach, the translations were evaluated using two prompts: a basic prompt and a detailed prompt tailored to Iranian cultural nuances. The results show that the human subtitler consistently aligned translations with Iranian cultural norms, adeptly navigating complex sensitivities. With the detailed prompt, ChatGPT and DeepSeek improved significantly in showed terms of religious and sociocultural accuracy, although only DeepSeek demonstrated moderate success with political content. Conversely, Gemini prioritized literal translations in sociocultural contexts regardless of prompt specificity. These findings emphasize the critical role of nuanced prompt engineering in AI-assisted translation, while affirming the unparalleled ability of human subtitlers to address cultural intricacies.