CoDeFont: One-Shot Font Generation via Multi-Scale Factor-Specific Representation Learning

Rejoy Chakraborty; Prasun Roy; Saumik Bhattacharya; Umapada Pal

cs.CVOct 15, 2026

CoDeFont: One-Shot Font Generation via Multi-Scale Factor-Specific Representation Learning

Authors: Rejoy Chakraborty, Prasun Roy, Saumik Bhattacharya, Umapada Pal

Organizations: Indian Statistical Institute Kolkata · Indian Institute of Technology Kharagpur

Abstract

One-shot Font Generation (OFG) aims to synthesize a target glyph by transferring the stylistic characteristics of a single reference glyph while preserving the structure of the target character. However, accurately capturing font style from a single reference image remains challenging due to the limited stylistic information available in a single glyph, and this challenge is further amplified for complex scripts. We propose CoDeFont, a script-agnostic OFG framework that addresses this challenge through multi-level style-content factorization and progressive style-content reconciliation. CoDeFont independently aggregates complementary style and structural information across multiple spatial levels to obtain factor-specific representations, while a contrastive objective explicitly encourages separation between style and content embeddings. During reconstruction, the disentangled representations are progressively injected through multi-level style conditioning, enabling stylistic information to be incorporated at different decoding resolutions while preserving the underlying glyph structure. We evaluate CoDeFont on structurally distinct Bengali and Latin scripts across seen and unseen font styles. Extensive quantitative and qualitative analyses demonstrate consistent improvements over existing state-of-the-art methods, highlighting the effectiveness of the proposed framework for OFG across diverse writing systems.

Figures & tables

Figure 1: Overview of the proposed CoDeFont . The generator independently encodes the style reference $x_{s}$ and content reference $x_{c}$ using the Style Encoder $E_{s}$ and Content Encoder $E_{c}$ to obtain style embedding $e_{s}$ and content embedding $e_{c}$ , respectively. The factorized representations are progressively reconciled in the Decoder $D_{sc}$ to generate the target glyph $\hat{y}$ . The Discriminator provides adversarial supervision and jointly performs auxiliary style and content classification, encouraging style consistency and content preservation.

Figure 2: Architectural overview of the proposed Style Encoder $E_{s}$ . Given a style image $x_{s}$ , the encoder extracts multi-scale style features through parallel convolutional branches with different kernel sizes and fuses them into a unified representation. The fused features are then processed by the Style-Query Attention Head (SQAH) to obtain the final style embedding $e_{s}$ .

Figure 3: Architecture of the proposed Content Encoder $E_{c}$ . Given a content image $x_{c}$ , the encoder extracts multi-scale features through successive convolutional blocks. The three downsampling features are processed by the Content Pooling Head (CPH) and adaptively fused using learnable scale weights to obtain the final content embedding $e_{c}$ .

Figure 4: Architecture of the proposed Decoder $D_{sc}$ . The decoder projects the content embedding into a low-resolution latent feature map and progressively upsamples it through the Multi-level Adaptive Style Fusion Module (MASFM) , which injects style information at multiple decoding levels using AdaIN [ 8 ] and channel-wise gating. The refined features are finally mapped to the target glyph $\hat{y}$ through a convolutional output layer.

Method	L1 $\downarrow$	RMSE $\downarrow$	SSIM $\uparrow$	LPIPS $\downarrow$	L1 $\downarrow$	RMSE $\downarrow$	SSIM $\uparrow$	LPIPS $\downarrow$	User Score $\uparrow$
	Unseen Fonts				Seen Fonts
Bengali
FANnet [ 20 ]	0.027	0.118	0.903	0.092	0.023	0.108	0.926	0.084	3.333
MA-Font [ 17 ]	0.032	0.148	0.893	0.071	0.030	0.145	0.896	0.069	1.795
PatchFont [ 14 ]	0.042	0.181	0.836	0.057	0.041	0.178	0.842	0.053	3.590
FontDiffuser [ 31 ]	0.020	0.094	0.912	0.047	0.014	0.067	0.954	0.043	43.846

Table 1: Quantitative comparison with existing SOTA methods on Unseen and Seen Bengali and Latin fonts. The top and bottom sections correspond to Bengali and Latin, respectively. The best and second-best results are highlighted in bold and underlined .

Figure 5: Qualitative comparison of our proposed method with existing SOTA techniques on Unseen and Seen Bengali and Latin fonts. The top and bottom sections correspond to Bengali and Latin, respectively. Red boxes highlight major failures, including poor style adaptation, structural deformation, and missing strokes, while blue boxes indicate minor artifacts and structural inconsistencies for existing methods.

Method	#Parameters $\downarrow$
FANnet [ 20 ]	4.2M
MA-Font [ 17 ]	34.4M
PatchFont [ 14 ]	51.3M
FontDiffuser [ 31 ]	323.2M
FASTER [ 5 ]	82.9M
Proposed	73.5M

Table 2: Comparative analysis of SOTA methods on the number of trainable parameters.

Figure 6: The t-SNE [ 25 ] visualization of the style ( $e_{s}$ ) and content ( $e_{c}$ ) embeddings, illustrating their factor-specific clustering.

$e_{s}$ by Style	$e_{s}$ by Content	$e_{c}$ by Style	$e_{c}$ by Content
0.695	-0.055	-0.017	0.999

Table 3: Silhouette scores of the style and content embeddings under different label assignments

Emb. Dim.	L1 $\downarrow$	RMSE $\downarrow$	SSIM $\uparrow$	LPIPS $\downarrow$
32	0.018	0.095	0.945	0.033
64	0.017	0.090	0.949	0.030
128	0.017	0.091	0.948	0.031
256	0.018	0.096	0.944	0.034
512	0.033	0.151	0.886	0.077

Table 4: Effect of embedding dimension on the performance of our proposed method. Best results are shown in bold .

Loss Function		Metrics
$\mathcal{L}_{\text{dist}}$	$\mathcal{L}_{\text{cls}}$	L1 $\downarrow$	RMSE $\downarrow$	SSIM $\uparrow$	LPIPS $\downarrow$
✗	✗	0.034	0.251	0.852	0.153
✗	✓	0.029	0.219	0.874	0.132
✓	✗	0.022	0.168	0.885	0.072
✓	✓	0.017	0.090	0.949	0.030

Table 5: Quantitative analysis of the effectiveness of

\mathcal{L}_{\text{cls}}

and

\mathcal{L}_{\text{dist}}

on the generation quality. Better results are shown in bold .

Gating	L1 $\downarrow$	RMSE $\downarrow$	SSIM $\uparrow$	LPIPS $\downarrow$
✗	0.019	0.097	0.938	0.038
✓	0.017	0.090	0.949	0.030

Table 6: Effect of gating mechanism on the performance of our proposed method. Better results are shown in bold .

MS Feature		Metrics
$E_{s}$	$E_{c}$	L1 $\downarrow$	RMSE $\downarrow$	SSIM $\uparrow$	LPIPS $\downarrow$
✗	✗	0.029	0.192	0.932	0.121
✗	✓	0.021	0.154	0.939	0.075
✓	✗	0.019	0.108	0.943	0.054
✓	✓	0.017	0.090	0.949	0.030

Table 7: Quantitative analysis of the effectiveness of multi-scale (MS) features in the style encoders

E_{s}

and

E_{c}

on generation quality. Best results are shown in bold .

Image Generation

Image generation studies models that create or edit images from text, examples, layouts, and other structured controls.

See all 474 papers

DRG-Font: Dynamic Reference-Guided Few-shot Font Generation via Contrastive Style-Content Disentanglement
Apr 15, 2026Rejoy Chakraborty, Prasun Roy, Saumik Bhattacharya +1Few-Shot Learning Conditional Generative Modeling
Indian Statistical Institute Kolkata · Indian Institute of Technology Kharagpur
Structure-Level Disentangled Diffusion for Few-Shot Chinese Font Generation
Feb 21, 2026Jie Li, Suorong Yang, Jian Zhao +1Few-Shot Learning Conditional Diffusion Models
State Key Laboratory for Novel Software Technology, Nanjing University, China · School of Artificial Intelligence, Nanjing University, China · Department of Computer Science and Technology, Nanjing University, China +1
SmartFont: Dynamic Condition Allocation for Few-Shot Font Generation
Jun 11, 2026Zian Yang, Zixin WangFew-Shot Learning Diffusion-Based Image Generation
Fudan University

Apr 15, 2026J/K move · Enter open · S save

DRG-Font: Dynamic Reference-Guided Few-shot Font Generation via Contrastive Style-Content Disentanglement

Rejoy Chakraborty, Prasun Roy, Saumik Bhattacharya, Umapada Pal

Indian Statistical Institute Kolkata · Indian Institute of Technology Kharagpur

Few-Shot Learning Conditional Generative Modeling Disentangled Representation Learning

Open paper AskPDF arXiv

Few-shot Font Generation aims to generate stylistically consistent glyphs from a few reference glyphs. However, capturing complex font styles from a few exemplars remains challenging, and the existing methods often struggle to retain discernible local characteristics in generated samples. This paper introduces DRG-Font, a contrastive font generation strategy that learns complex glyph attributes by decomposing style and content embedding spaces. For optimal style supervision, the proposed architecture incorporates a Reference Selection (RS) Module to dynamically select the best style reference from an available pool of candidates. The network learns to decompose glyph attributes into style and shape priors through a Multi-scale Style Head Block (MSHB) and a Multi-scale Content Head Block (MCHB). For style adaptation, a Multi-Fusion Upsampling Block (MFUB) produces the target glyph by combining the reference style prior and target content prior. The proposed method demonstrates significant improvements over state-of-the-art approaches across multiple visual and analytical benchmarks.

Abstract

Image Generation

Explore similar work

DRG-Font: Dynamic Reference-Guided Few-shot Font Generation via Contrastive Style-Content Disentanglement

Structure-Level Disentangled Diffusion for Few-Shot Chinese Font Generation

SmartFont: Dynamic Condition Allocation for Few-Shot Font Generation