Stroke Modeling Enables Vectorized Character Generation with Large Vectorized Glyph Model

2025-11-14

AI Summary

This paper addresses the challenges of structural modeling and controllability in vector-based Chinese character generation. We propose the Large Vector Glyph Model (LVGM), which decomposes characters into ordered stroke sequences, encodes strokes as discrete latent variables via stroke embeddings, and employs fine-tuned DeepSeek large language models for stroke-level autoregressive prediction. This work pioneers the deep integration of stroke-level modeling with large language models, enabling full-vector glyph synthesis from partial strokes, semantically coherent multi-character words, and original classical poetry. To support this, we construct the first large-scale Chinese SVG glyph dataset (900K samples). Experiments demonstrate LVGM’s strong scalability; expert evaluation confirms high geometric fidelity, artistic plausibility, and precise stroke-level editability. LVGM establishes a novel paradigm for dynamically editable, controllable Chinese font generation.

Technology Category

• Natural Language Processing: (Large) Language Models
• Machine Learning: Large Multimodal Models (LMMs)
• Computer Vision: Large Vision Models

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

Vectorized glyphs are widely used in poster design, network animation, art display, and various other fields due to their scalability and flexibility. In typography, they are often seen as special sequences composed of ordered strokes. This concept extends to the token sequence prediction abilities of large language models (LLMs), enabling vectorized character generation through stroke modeling. In this paper, we propose a novel Large Vectorized Glyph Model (LVGM) designed to generate vectorized Chinese glyphs by predicting the next stroke. Initially, we encode strokes into discrete latent variables called stroke embeddings. Subsequently, we train our LVGM via fine-tuning DeepSeek LLM by predicting the next stroke embedding. With limited strokes given, it can generate complete characters, semantically elegant words, and even unseen verses in vectorized form. Moreover, we release a new large-scale Chinese SVG dataset containing 907,267 samples based on strokes for dynamically vectorized glyph generation. Experimental results show that our model has scaling behaviors on data scales. Our generated vectorized glyphs have been validated by experts and relevant individuals.

Problem

• Generating vectorized Chinese characters from stroke sequences
• Predicting next stroke embeddings using fine-tuned language models
• Creating scalable vector graphics for typography and design applications

Innovation

• Stroke modeling enables vectorized character generation
• Fine-tuned LLM predicts next stroke embeddings
• Large-scale SVG dataset supports dynamic glyph generation