OnomatoGen: Onomatopoeia Generation with the Alpha-Channel in Manga

2025-09-29

AI Summary

This paper addresses the long-overlooked problem of onomatopoeia visual stylization in manga generation. We propose OnomatoGen, the first end-to-end model specifically designed for onomatopoeia generation and stylistic control. Unlike generic text style transfer methods, OnomatoGen explicitly models semantic-visual alignment between onomatopoeic text and scene emotion intensity by leveraging the alpha channel to govern shape, scale, and spatial placement of onomatopoeic glyphs. The framework integrates three key components: (1) fine-grained textual semantic modeling, (2) cross-modal visual–semantic alignment, and (3) alpha-channel–driven controllable image generation. Quantitative and qualitative evaluations demonstrate that OnomatoGen significantly outperforms existing approaches in both visual expressiveness and contextual coherence of generated onomatopoeia. It establishes a novel paradigm for AI-driven manga synthesis, enabling semantically grounded, stylistically adaptive, and compositionally aware onomatopoeia rendering.

Technology Category

• Natural Language Processing: Generation
• Computer Vision: Generative Adversarial Networks (GANs) for Vision
• Cognitive Modeling & Cognitive Systems: Computational Creativity

Application Category

• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications
• Semantics and Knowledge: Scalable techniques for the creation, curation, publication, maintenance, and consumption of large, Web-based, structured, reusable, knowledge graphs and ontologies
• Web Mining and Content Analysis: Web data generation and simulation

Abstract

Onomatopoeia is an important element for textual messaging in manga. Unlike character dialogue in manga, onomatopoeic expressions are visually stylized, with variations in shape, size, and placement that reflect the scene's intensity and mood. Despite its important role, onomatopoeia has not received much attention in manga generation. In this paper, we focus on onomatopoeia generation and propose OnomatoGen, which stylizes plain text to an onomatopoeic style. We empirically show the unique properties of onomatopoeia generation, which differ from typical text stylization methods, and that OnomatoGen can effectively stylize plain text in an onomatopoeia style.

Problem

• Generating visually stylized onomatopoeia for manga
• Transforming plain text into expressive sound representations
• Addressing unique challenges in manga text stylization

Innovation

• Generates manga onomatopoeia from plain text
• Stylizes text with visual shape and size variations
• Uses Alpha-Channel for manga-specific text representation