[Paper] LottieGPT: Tokenizing Vector Animation for Autoregressive Generation
Paper Overview
- Research area: cs.CV
- Authors: Junhao Chen, Kejun Gao, Yuehan Cui, Mingze Sun, Mingjin Chen, Shaohui Wang, Xiaoxiao Long, Fei Ma, Qi Tian, Ruqi Huang, Hao Zhao
- Published: 2026-04-13
- arXiv: 2604.11792
- Lottie Tokenizer: encodes hierarchical geometry, transforms, and keyframe motion as compact, semantic token sequences
- LottieAnimation-660K: a large-scale dataset for vector animation generation
- LottieGPT: a Qwen-VL-based autoregressive model producing editable vector animations from text or visual prompts
Summary
Despite rapid progress in video generation, existing models are incapable of producing vector animation, a dominant and highly expressive form of multimedia on the Internet. Vector animations offer resolution-independence, compactness, semantic structure, and editable parametric motion representations, yet current generative models operate exclusively in raster space and thus cannot synthesize them.This paper presents the first framework for tokenizing and autoregressively generating vector animation. Building on Lottie, a widely deployed JSON-based animation standard, the authors design a custom Lottie Tokenizer that encodes hierarchical geometric primitives, transforms, and keyframe-based motion into compact, semantically aligned token sequences. They further construct the LottieAnimation-660K dataset and fine-tune Qwen-VL to create LottieGPT, a model capable of generating coherent, editable vector animations directly from natural language or visual prompts.