HairDeck: Animation-Ready Stylized Hair Cards from a Single Image

Yuze He, Yixiang Wei, Wang Zhao, Chuhao Chen, Yanning Zhou, Yong-Jin Liu

Tsinghua University · Carnegie Mellon University · The Chinese University of Hong Kong

Published October 29, 2026 · on the feed October 30, 2026 · ▲ 41 · arXiv 2610.40599

3D, worlds & gamesImage generation & editing Full code arXiv ↗Hugging Face ↗Weights ↗

The abstract

The hair of stylized characters in games and avatars is built from hair cards: artist-placed ribbons whose geometry is the visible hair, animated by spring bones. Existing image-based methods produce realistic hair or a single fused mesh, while CHARM, the only generator of stylized hair cards, decodes a hairstyle autoregressively as a long sequence of control points. Observing that a card hairstyle is an unordered set of artist-defined primitives, we propose HairDeck, which generates it as one: each card is a single token, and a set diffusion transformer denoises all cards together. From a single image, HairDeck generates the cards and the spring-bone chains that drive them, yielding an animatable asset that exports to ready-to-use downstream formats. On CHARM's own test set, it lowers Chamfer distance by 44% and card-orientation error by more than half while running 33 times faster, and from single images it outperforms CHARM, image-to-3D pipelines and strand-based methods. Project page: https://hyzcluster.github.io/HairDeck

The Brief

1:53

A short podcast summary of the paper. Written and voiced by AI.

Transcript

HairDeck. Animation-Ready Stylized Hair Cards from a Single Image. By Yuze He and five others, from Tsinghua University, Carnegie Mellon University, and one other institution. Published on the twenty-ninth of October.

Stylized characters in games and avatars get their hair from hair cards: ribbons placed by artists, whose geometry is the visible hair, animated by spring bones. Existing image-based methods produce either realistic hair or a single fused mesh, neither of which is a set of cards. CHARM, the only generator of stylized hair cards, decodes a hairstyle autoregressively, as a long sequence of control points. A hairstyle in that framing is one long string, produced token after token, even though the cards that make it up are separate pieces. And a usable asset needs more than shape: it needs the bones that make the cards move.

HairDeck starts from an observation. A card hairstyle is an unordered set of artist-defined primitives, so it can be generated as a set. Each card becomes a single token, and a set diffusion transformer denoises all the cards together, instead of emitting them one point at a time.

From a single image, HairDeck generates both the cards and the spring-bone chains that drive them. The result is an animatable asset that exports to ready-to-use downstream formats. That closes the gap between a generated shape and something a game or avatar pipeline can use directly: the output is already in the form artists build, cards plus the spring bones that animate them, rather than a mesh that would still have to be converted and rigged.

On CHARM's own test set, HairDeck lowers Chamfer distance by forty-four percent, cuts card-orientation error by more than half, and runs thirty-three times faster. Working from single images, it outperforms CHARM, image-to-3D pipelines, and strand-based methods.

The gain rests on matching the representation to the object. A hairstyle made of cards has no natural order, and treating it as a set, one token per card, is what lets the whole thing be generated at once and rigged for motion.

More with an account

Listen to every day’s papers

Daily playlists, deep dives, quests for the papers, and quizzes on their key terms.

Join free
KipQuest: the day’s AI research, as podcasts and quests. All papers Android app Privacy Contact Episodes are written and voiced by AI from each paper. Summaries can contain mistakes: the paper is the source.