Abstract
Generative video systems can produce generic or incoherent output when given vague user prompts. A creative rewriter (CR) system may function as an automated script-generation system that can transform high-level user intent into a comprehensive, structured script. The CR system is configured to analyze minimal inputs, such as a short text phrase or a static image, and can generate detailed instructions specifying elements like plot, narrative arc, camera cinematography, pacing, character actions, and sound design. This process can provide a downstream video synthesis model with a detailed blueprint, which may improve the generation of narratively coherent and cinematically detailed video content from simple prompts. This approach can also reduce reliance on a user's ability to articulate complex directorial commands.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Sluzhaev, Evgeny; Weisz, Ágoston; Bulski, Aleksander; Fadeeva, Asya; Vernikos, Giorgos; and Federico Xu, Xingyu, "Transforming High-Level Intent into Detailed Directorial Scripts for Generative Video", Technical Disclosure Commons, (May 11, 2026)
https://www.tdcommons.org/dpubs_series/10057