Text-Conditioned Diffusion Model for High-Fidelity Korean Font Generation
Published in 2025 International Conference on Information Networking (ICOIN), 2025
DK-Font presents a one-shot, text-conditioned diffusion model for generating high-fidelity Korean fonts. By fusing a phonetic-aware text encoder with a perceptual loss, the method preserves stroke integrity and style consistency across all 2,350 Hangul characters, boosting SSIM from 0.812 to 0.857 and lowering FID to 10.45 compared with Diff-Font. Few-shot capabilities allow synthesizing full font sets from just 3–5 reference glyphs.
Recommended citation: Abdul Sami, Jaeyong Choi. (2025). "Text-Conditioned Diffusion Model for High-Fidelity Korean Font Generation." 2025 International Conference on Information Networking (ICOIN).
Download Paper