“From Style to Identity: AI Pipelines for Visual and Character Coherence in Film” by Zhang
Conference:
Type(s):
Title:
- From Style to Identity: AI Pipelines for Visual and Character Coherence in Film
Session/Category Title:
- Images, Video & Computer Vision
Presenter(s)/Author(s):
Abstract:
We introduce a modular, open-source pipeline that combines multiple custom-trained LoRA and ControlNet models to disentangle style and identity, enabling fast, visually and narratively consistent AI-generated short films，validated through two award-winning multi-scene productions.
References:
[1] ComfyUI. 2025. ComfyUI. GitHub repository. Retrieved from github.com/comfyanonymous/ComfyUI
[2 Edward J. Hu, Yelong Shen, Phil Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Lu Wang, and Weizhu Chen. 2021. LoRA: Low-Rank Adaptation of Large Language Models. arXiv preprint arXiv:2106.09685.
[3] Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Björn Ommer. 2022. High-resolution image synthesis with latent diffusion models. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). 10684–10695.
