The challenge of automatically synthesizing high-quality vector fonts, par-ticularly for writing systems (e.g., Chinese) consisting of huge amounts ofcomplex glyphs, remains unsolved. Existing font synthesis techniques fallinto two categories: 1) methods that directly generate vector glyphs, and2) methods that initially synthesize glyph images and then vectorize them.However, the first category often fails to construct complete and correctshapes for complex glyphs, while the latter struggles to efficiently synthesizehigh-resolution (i.e., 1024x1024 or higher) glyph images while preservinglocal details. In this paper, we introduce HFH-Font, a few-shot font synthe-sis method capable of efficiently generating high-resolution glyph imagesthat can be converted into high-quality vector glyphs. More specifically,our method employs a diffusion model-based generative framework with component-aware conditioning to learn different levels of style informationadaptable to varying input reference sizes. We also design a distillationmodule based on Score Distillation Sampling for 1-step fast inference, and astyle-guided super-resolution module to refine and upscale low-resolutionsynthesis results. Extensive experiments, including a user study with profes-sional font designers, have been conducted to demonstrate that our methodsignificantly outperforms existing font synthesis approaches. Experimentalresults show that our method produces high-fidelity, high-resolution rasterimages which can be vectorized into high-quality vector fonts. Using ourmethod, for the first time, large-scale Chinese vector fonts of a quality com-parable to those manually created by professional font designers can beautomatically generated.
