CineVision: An Interactive Pre-visualization Storyboard System for Director–Cinematographer Colla...
summary
CineVision is an AI-driven pre-visualization system that links script input to real-time storyboard images, using fine-tuned Stable Diffusion on a MovieNet-derived shot dataset and ChatGPT-4–written style notes for ten director emulations. It offers dynamic lighting controls, consistent character appearance across shots, film-database matching, and a hierarchical prompt UI. A 24-participant lab study (12 film experts, 12 non-experts) reports shorter task times, lower workload, higher usability and collaboration scores versus DALL·E 3 and hand-drawn baselines; it performs best for dialogue-heavy scenes and is less helpful for complex action sequences. Code and demo: github.com/TonyHongtaoWu/CineVision.
AI-written summary · about
chapters (11)
- 0:00Introduction to CineVision
- 1:49Design challenges and goals
- 2:52System pipeline overview
- 3:53Lighting and style variation
- 5:35Character consistency
- 6:41Film database matching
- 7:22Demo workflow
- 8:47Real-time collaboration
- 9:09Evaluation setup
- 10:34Summary and future work
- 10:54Q&A session
related talks
29mGenerative AI for Engaging User Storyboards
01h 15mPre-sets and Post-sets: Poetics and Projections
12mAI Content Generation
18mDreamina Seedance 2.0
29mWhere design meets story
33mVisual thinking - simplify complex
17mStoryPrint: an interactive visualization of stories
39mVideo Generation + AI: From Production to Product