Scholarlink

The rapid development of Diffusion models and the declarative nature of interfaces developed for the public require automation methods, where media production can harness natural language as a mode of representation but not necessarily of interaction with humans. This article describes an image-to-video Diffusion system which removes practitioners from the process of defining prompts when producing images with conditional reference, documenting a set of results with a custom dataset of oil paintings. Our research focuses on the appropriation of trained model ensembles that are coordinated to produce indefinite sets of frames with occasional human intervention utilising timeline-based architectures. The proposed system automates a CLIP-guided DDPM with a supplementary depth estimation model and through a set of compositing techniques we found that results with coincidental and diverging descriptions can be useful for moving-image element composition. Our experiments focus on the representation of human figure and its morphological transformation.

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

All YIN No YANG: Geometric Abstraction of Oil Paintings with Trained Models, Noise and Self-reference

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Luís Arandas,
  • Iulia Ionescu,
  • Murad Khan,
  • Mick Grierson,
  • Miguel Carvalhais

摘要

The rapid development of Diffusion models and the declarative nature of interfaces developed for the public require automation methods, where media production can harness natural language as a mode of representation but not necessarily of interaction with humans. This article describes an image-to-video Diffusion system which removes practitioners from the process of defining prompts when producing images with conditional reference, documenting a set of results with a custom dataset of oil paintings. Our research focuses on the appropriation of trained model ensembles that are coordinated to produce indefinite sets of frames with occasional human intervention utilising timeline-based architectures. The proposed system automates a CLIP-guided DDPM with a supplementary depth estimation model and through a set of compositing techniques we found that results with coincidental and diverging descriptions can be useful for moving-image element composition. Our experiments focus on the representation of human figure and its morphological transformation.