Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Generative AI for Cel-Animation: A Survey

Yolo Y. Tang, Junjia Guo, Pinxin Liu, Zhiyuan Wang, Hang Hua, Jia-Xing Zhong, Yunzhong Xiao, Chao Huang, Luchuan Song, Susan Liang, Yizhi Song, Liu He, Jing Bi, Mingqian Feng, Xinyang Li, Zeliang Zhang, Chenliang Xu

TL;DR

This survey analyzes how Generative AI can transform Cel-Animation by mapping LLMs, multimodal models, and diffusion-based techniques to each stage of the traditional pipeline—pre-production, production, and post-production. It highlights concrete tools (e.g., AniDoc, ToonCrafter, AniSora) and case studies (The Dog & The Boy, Rock, Paper, Scissors, Twins Hinahima) that demonstrate reductions in manual labor and enhanced creative exploration, while detailing challenges in stylistic consistency, temporal coherence, and ethical considerations. The paper also discusses technical approaches for script, setting, storyboard generation, layout and keyframe generation, inbetweening, colorization, and post-production tasks, along with data governance, licensing, and scalability issues. Its findings emphasize that GenAI can democratize animation workflows and expand creative possibilities, provided that robust control interfaces, provenance, and alignment with artistic intent are addressed in future work.

Abstract

Traditional Celluloid (Cel) Animation production pipeline encompasses multiple essential steps, including storyboarding, layout design, keyframe animation, inbetweening, and colorization, which demand substantial manual effort, technical expertise, and significant time investment. These challenges have historically impeded the efficiency and scalability of Cel-Animation production. The rise of generative artificial intelligence (GenAI), encompassing large language models, multimodal models, and diffusion models, offers innovative solutions by automating tasks such as inbetween frame generation, colorization, and storyboard creation. This survey explores how GenAI integration is revolutionizing traditional animation workflows by lowering technical barriers, broadening accessibility for a wider range of creators through tools like AniDoc, ToonCrafter, and AniSora, and enabling artists to focus more on creative expression and artistic innovation. Despite its potential, challenges like visual consistency, stylistic coherence, and ethical considerations persist. Additionally, this paper explores future directions and advancements in AI-assisted animation. For further exploration and resources, please visit our GitHub repository: https://github.com/yunlong10/Awesome-AI4Animation

Generative AI for Cel-Animation: A Survey

TL;DR

This survey analyzes how Generative AI can transform Cel-Animation by mapping LLMs, multimodal models, and diffusion-based techniques to each stage of the traditional pipeline—pre-production, production, and post-production. It highlights concrete tools (e.g., AniDoc, ToonCrafter, AniSora) and case studies (The Dog & The Boy, Rock, Paper, Scissors, Twins Hinahima) that demonstrate reductions in manual labor and enhanced creative exploration, while detailing challenges in stylistic consistency, temporal coherence, and ethical considerations. The paper also discusses technical approaches for script, setting, storyboard generation, layout and keyframe generation, inbetweening, colorization, and post-production tasks, along with data governance, licensing, and scalability issues. Its findings emphasize that GenAI can democratize animation workflows and expand creative possibilities, provided that robust control interfaces, provenance, and alignment with artistic intent are addressed in future work.

Abstract

Traditional Celluloid (Cel) Animation production pipeline encompasses multiple essential steps, including storyboarding, layout design, keyframe animation, inbetweening, and colorization, which demand substantial manual effort, technical expertise, and significant time investment. These challenges have historically impeded the efficiency and scalability of Cel-Animation production. The rise of generative artificial intelligence (GenAI), encompassing large language models, multimodal models, and diffusion models, offers innovative solutions by automating tasks such as inbetween frame generation, colorization, and storyboard creation. This survey explores how GenAI integration is revolutionizing traditional animation workflows by lowering technical barriers, broadening accessibility for a wider range of creators through tools like AniDoc, ToonCrafter, and AniSora, and enabling artists to focus more on creative expression and artistic innovation. Despite its potential, challenges like visual consistency, stylistic coherence, and ethical considerations persist. Additionally, this paper explores future directions and advancements in AI-assisted animation. For further exploration and resources, please visit our GitHub repository: https://github.com/yunlong10/Awesome-AI4Animation
Paper Structure (41 sections, 3 equations, 6 figures)

This paper contains 41 sections, 3 equations, 6 figures.

Table of Contents

  1. Introduction
  2. The Handcrafted Cel Age (1920s-2010s)
  3. The Computer-Assisted Cel Age (1980s-present).
  4. The AIGC Cel Era (2020s onward).
  5. Preliminary
  6. Conventional Cel-Animation Pipeline
  7. Pre-production
  8. Production
  9. Post-production
  10. Other Processes
  11. Generative AI
  12. Large Language Models (LLMs)
  13. GAN, VAE, and Diffusion Models
  14. GenAI for Cel-Animation
  15. GenAI for Pre-production
  16. ...and 26 more sections

Figures (6)

  • Figure 1: The three major phases in Cel-Animation history: the Handcrafted Cel Age (1920s-2010s), the Computer-Assisted Cel Age (1980s-present), and the emerging AIGC Cel Era (2020s onward). A layered structure of Cel-Animation is also shown.
  • Figure 2: Workflow diagram of a traditional animation pipeline.
  • Figure 3: The taxonomy of GenAI for Cel-Animation is primarily organized by the tasks involved in the production workflow. Key steps are highlighted by Green, Orange, and Purple. Methods capable of multiple tasks within the Cel-Animation production are indicated by Red, while Blue signifies methods that are not originally developed for Cel-related tasks but have the potential to be applied to Cel-Animation.
  • Figure 4: Examples of AI-generated Settings with Midjourney midjourney and Stable Diffusion rombach2022high, showing various designs including items, characters, scenes, and mecha.
  • Figure 5: Comparison of colorization generated by GenAI models. The example is adapted from meng2024anidoc.
  • ...and 1 more figures