现代电影技术Issue(7)：4-13,10.DOI:10.3969/j.issn.1673-3215.2026.07.001
面向电影前期视觉开发的图像智能生成框架研究
Research on intelligent image generation framework for film pre-production vi-sual development
摘要
Abstract
In film development,storyboards,concept art,and art assets often require multiple rounds of iteration around character design,prop style,scene atmosphere,and shot composition.Existing global semantic editing and mask-based repainting methods still struggle to balance structural stability and semantic independence in multi-region collaborative edit-ing.To address this issue,this paper constructs ReMAP,an intelligent image generation framework for film pre-production visual development.The proposed method decomposes complex storyboards into independent semantic asset regions and establishes explicit constraint mappings between spatial masks and corresponding prompts,enabling precise local control of film art assets and stable preservation of non-edited regions,while effectively supporting multi-region collaborative re-painting and multi-variant creative comparison.To reduce the cost of prompt engineering in complex scenes,this paper fur-ther constructs a multimodal standard operating procedure prompt generation mechanism,which adaptively extracts visual semantic features from reference assets and original storyboards and automatically generates reusable scene description templates containing global style and regional constraints.Experimental results show that,under the unified Flux2-Klein backbone,ReMAP outperforms baselines such as Prompt-to-Prompt global semantic editing and DiffEdit local mask edit-ing in multi-region semantic isolation,asset structural consistency,and non-edited region stability,providing low-cost and highly interactive intelligent technical support for film art design and storyboard production.关键词
扩散模型/人工智能生成内容(AIGC)/图像编辑/可控图像生成Key words
Diffusion Model/AI-generated Content(AIGC)/Image Editing/Controllable Image Generation分类
信息技术与安全科学引用本文复制引用
黄东晋,张诗雨,曲建涛,王倩..面向电影前期视觉开发的图像智能生成框架研究[J].现代电影技术,2026,(7):4-13,10.基金项目
上海市人才发展资助项目(2021016) （2021016）
上海市教育科学研究项目"生成式人工智能赋能影视创制人才培养的实践研究"(GSC2026099) （GSC2026099）
中国电影地缘文化研究基地2025年度项目"国产中小成本电影的地缘版图研究"(25Z0105). （25Z0105）
