Manga is a comic drawn by black and white paints gaining popularity around the world. Onomatopoeia in Manga specifically appeals to the audience with its unique visual styles, which convey sound, motion, and emotion. Visual onomatopoeia translation requires the clean replacement of Japanese onomatopoeia with onomatopoeia in the other language while preserving their visual style. Existing approaches often produce residual artifacts or style inconsistency when removing the Japanese onomatopoeia and rendering stylized English onomatopoeia. To approach these problems, we present OnomatoBridge, a filtering pipeline for visual onomatopoeia translation. We evaluate OnomatoBridge from Japanese to English on the Manga109 onomatopoeia dataset and compare it with baseline image editing models. Experimental results show that the filtered outputs by the proposed method outperform those of conventional methods. OnomatoBridge improves English text correctness by roughly 10 to 25 points and reduces residual Japanese text by about 20 to 50% in relative terms.
Figures & tables
Figure 1. Left: Onomatopoeia appearing in Manga. From DualJustice, c◯ Yusuke Takeyama, Arisa c◯ Ken Yagami. Right: Our target task, which replaces Japanese onomatopoeia with the English one. From AppareKappore c◯ Hirsohi Kanno, GarakutayaManta c◯ Tatsuki Nouda
Figure 2. Each component of OnomatoBridge. Right: Whereas the input image of OnomatoBridge is processed to inpaint the onomatopoeia, it is simultaneously processed to select the best extracted onomatopoeia. Left: The output of the onomatopoeia extractor is transformed to the English onomatopoeia and it is pasted to the output of inpainter. From EvaLady c◯ Mio Iisen
Model name
NED ↑
JPDet rate ↓
Style ↑
Residual ↑
Readability ↑
Naïve baseline
0.8918
0.5108
-
-
-
Flux.1 Kontext ( Labs et al., 2025 )
0.5391
0.2772
2.151
3.548
3.014
Qwen Image Edit ( Wu et al., 2025 )
0.5673
0.4547
1.986
2.548
2.863
Gemini Image 3 pro
0.7182
0.3949
2.945
1.944
3.945
OnomatoBridge (ours)
0.7826
0.2246
3.098
3.946
4.121
Table 1. Quantitative comparison and user study results. NED (Normalized Edit Distance) measures text accuracy, while JPDet rate (Japanese Detection Rate) indicates the ratio of original Japanese text remaining after editing. Style, Residual, and Readability are subjective scores (1–5) evaluating artistic integration, background inpainting quality, and text legibility, respectively.
Figure 3. Quantitative comparison between OnomatoBridge and baselines. OnomatoBridge reflects the original style of Japanese onomatopoeia and cleanly removes the residuals.
In manga translation, visual context outside speech-bubble text, including speaker identities, referents, and scene tone, is depicted in page images, making it natural to refer to images during translation evaluation. However, most conventional manga translation evaluations rely solely on text-only automatic metrics, leaving the impact of image input insufficiently explored. We propose a practical validation framework that employs the same vision-language model (VLM) as both a text-only judge (TOJ) and an image-aware judge (IAJ) to investigate evaluation divergences associated with image-aware evaluation procedures. We use a Japanese-English manga dataset of official translations aligned at the page and speech-bubble levels. First, adding page images to TOJ improved the primary VLM's detection of deliberately injected, visually dependent degradations. Second, we analyzed evaluation divergences between TOJ and IAJ, finding different error counts and error-category distributions. Human verification did not reveal a consistent preference between TOJ and IAJ.
Manga is a culturally distinctive multimodal medium and one of the most influential forms of Japanese popular culture. As AI systems increasingly target manga understanding, OCR, and translation, Manga109 has become a foundational dataset for manga-related AI research. However, the current Manga109 dataset contains inaccurate transcriptions and coarse annotations, which do not align well with modern OCR and multimodal manga understanding tasks. In this work, we revisit the dialogue text annotations of Manga109 and identify five categories of annotation issues, including inaccurate transcriptions, missing text regions, overlapping dialogue and onomatopoeia, and under-segmented speech balloons. To address these issues, we combine OCR-based issue detection and manual revision to construct Manga109-v2026, revising approximately 29,000 dialogue annotations. Our revisions better align Manga109 with modern OCR and multimodal manga understanding systems while preserving expressive structures characteristic of manga.
Jeonghun Baek, Atsuyuki Miyai, Shota Onohara +2
The University of Tokyo, Tokyo, Japan · Mantra Inc., Tokyo, Japan
Kamon (family crests) are an important part of Japanese culture and a natural test case for compositional visual recognition: each crest combines a small number of symbolic choices, but the space of possible descriptions is sparse. We introduce KamonBench, a grammar-based image-to-structure benchmark with 20,000 synthetic composite crests and auxiliary component examples. Each composite crest is paired with a formal kamon description language (kamon yogo) description, a segmented Japanese analysis, an English translation, and a non-linguistic program code. Because each synthetic crest is generated from known factors, namely container, modifier, and motif, KamonBench supports evaluation beyond caption-level accuracy: direct program-code factor metrics, controlled factor recombination splits, counterfactual motif-sensitivity groups under fixed container-modifier contexts, and linear probes of factor accessibility. We include baseline results for a ViT encoder with a Transformer decoder, two VGG n-gram decoders with and without learned positional masks, and fine-tuned Qwen3-VL-4B and Qwen3-VL-8B models. KamonBench therefore provides a controlled testbed for sparse compositional visual recognition and factor recovery in vision-language models.