VecFontLLM: Anchor-Guided Direct Synthesis of Chinese Vector Fonts

2026-07-19

AI Summary

Existing methods struggle to directly generate Chinese vector fonts with intricate structures and entangled multi-scale features, often relying on rasterization-based post-processing and underperforming on complex glyphs. This work proposes VecFontLLM, an anchor-guided multimodal large language model that enables few-shot, end-to-end vector font synthesis by first predicting and refining an anchor skeleton to establish component layout and then completing Bézier control points to recover curvature and stylistic details. Notably, this approach is the first to produce high-quality Chinese vector fonts without any rasterization or vectorization post-processing. By introducing a confidence-driven generation chain, it significantly enhances stability on complex characters. Experiments demonstrate that VecFontLLM achieves rendering quality comparable to raster-domain models, yields vector command distributions closely aligned with real fonts, and consistently outperforms existing methods across multiple benchmarks.

Abstract

Direct generation of Chinese vector fonts is a challenging and ongoing problem. A Chinese vector glyph contains complex component structure, anchor layout, and Bézier curve details, which work at different scales, but a standard vector sequence writes them together in one long sequence, making the task of vector font synthesis challenging. Existing direct vector generators often fail on complex characters, while raster-domain methods must vectorize the synthesized glyph images afterward. To address the above-mentioned problem, this paper proposes VecFontLLM, an anchor-guided multimodal large language model for direct few-shot synthesis of Chinese vector fonts. Our key idea is to generate vector glyphs through anchors rather than a standard vector sequence. Specifically, the proposed VecFontLLM first predicts and refines an anchor scaffold that fixes the coarse layout of components and contours, and then completes Bézier control points to recover local curvature and style. At test time, a confidence-guided generation chain samples multiple component candidates and continues synthesis from the highest-confidence one, improving stability for complex glyphs. This work demonstrates, for the first time, high-quality few-shot synthesis of complex Chinese vector glyphs directly in the vector domain, without raster generation or vectorization. Experiments on several Chinese font datasets show substantial improvements over existing vector font synthesis methods, competitive glyph rendering quality against raster-domain baselines, and vector command distributions close to real fonts.

Problem

• Chinese vector fonts
• direct synthesis
• vector font generation
• complex glyph structure
• anchor layout

Innovation

• anchor-guided generation
• vector font synthesis
• few-shot learning
• Bezier curve modeling
• multimodal LLM