The film industry is a major contributor to the economy and provides thousands of jobs. One of its biggest challenges is generating reallistic set designs, a time-consuming and costly process for production teams. Period films, in particular, require historically accurate and visually coherent sets, increasing production costs. This study leverages Generative AI techniques to create historically inspired set design visuals directly from film scripts. Large Language Models (LLMs), such as LLaMA 3.2, extract keywords from scripts, which are then passed to a fine-tuned Stable Diffusion model for generating Mughal architectural designs. The generated images are evaluated using CLIP scores, showing an improvement of 3.37 points over the baseline model’s 32.8. Results highlight AI’s potential in making period filmmaking more efficient, cost-effective, and accessible while maintaining artistic and historical fidelity.

错误:搜索内容不能为空,请输入英文关键词
错误:关键词超出字数限制,请精简
高级检索

Cinematic Vision: Set Design for Period Films from Script-to-Scene Using LLMs

  • Atla Srivatsav,
  • Sujala Achanta,
  • Dhruv Surti,
  • Namrata Hangala,
  • D. Uma

摘要

The film industry is a major contributor to the economy and provides thousands of jobs. One of its biggest challenges is generating reallistic set designs, a time-consuming and costly process for production teams. Period films, in particular, require historically accurate and visually coherent sets, increasing production costs. This study leverages Generative AI techniques to create historically inspired set design visuals directly from film scripts. Large Language Models (LLMs), such as LLaMA 3.2, extract keywords from scripts, which are then passed to a fine-tuned Stable Diffusion model for generating Mughal architectural designs. The generated images are evaluated using CLIP scores, showing an improvement of 3.37 points over the baseline model’s 32.8. Results highlight AI’s potential in making period filmmaking more efficient, cost-effective, and accessible while maintaining artistic and historical fidelity.