An ONNX Vision‑Encoder‑Decoder Manga OCR model built on kha‑white’s manga‑ocr and jzhang533’s 2025 baseline, supporting vertical/horizontal Japanese text, furigana, overlays and low‑quality images, trained on manga109‑s and synthetic data, and deployable via Hugging Face Optimum with TrOCRProcessor and ORTModelForVision2Seq.
Description in other languages
Statistics
Statistics as of Nov 16, 2026
Hugging Face downloads (last 30 days)
9,000
Hugging Face likes
10
Listed
Oct 7, 2025
Downloads (30-day window) over time
Show as a table
| Month | DL (30d) |
|---|---|
| Dec 2025 | — |
| Jan 2026 | — |
| Feb 2026 | — |
| Mar 2026 | — |
| Apr 2026 | — |
| May 2026 | — |
| Jun 2026 | — |
| Jul 2026 | 1,000 |
| Aug 2026 | 3,000 |
| Sep 2026 | 4,000 |
| Oct 2026 | 6,000 |
| Nov 2026 | 9,000 |
Downloads over the 30 days before the last snapshot of each month.
Popular in the same category
- manga-ocr-basekha-white299kManga OCR is a Vision Encoder‑Decoder OCR tool that reads vertical and horizontal Japanese manga text—including furigana—across diverse fonts and low‑quality images, with the source code freely available.
- meiki.txt.recognition.v0rtr4642kMeikiocr’s meiki.text.recognition.v0—a D‑FINE-based MobileNetV4 model fine‑tuned on Japanese video‑game text—delivers state‑of‑the‑art accuracy and latency for horizontal text by detecting up to 48 characters from 960×32 inputs, outputting each character with its bounding box and confidence score.
- meiki.text.detect.v0rtr4622kmeikiocr supplies a D‑FINE‑based, open‑weight text‑detection model for video‑games (v0.1 with MobileNet‑v4 backbones, two resolution variants and a 64‑box limit) and experimental low‑latency tiny and small variants trained on Japanese video‑games and manga.
- manga-ocrmayocream2kManga OCR is a Vision Encoder‑Decoder system that delivers high‑quality OCR of Japanese manga—including vertical and horizontal text with furigana overlays—across diverse fonts and low‑quality images, and can also be used for general printed Japanese OCR.
- sarashina2.2-vision-3bsbintuitions1kSarashina2.2‑Vision‑3B is a Japanese large vision‑language model (3 B parameters) built on Sarashina2.2‑3B‑Instruct and SigLIP’s image encoder, achieving competitive scores on Japanese VQA benchmarks such as BussinessSlide, Heron‑Bench, JDocQA, and JMMMU while also performing well on English tasks like DocVQA, InfoVQA, and RealworldQA.
- sarashina2.2-ocrsbintuitions1kSarashina2.2‑OCR is a 3‑B parameter, end‑to‑end OCR model that, after human‑preference optimization, parses Japanese and English documents into Markdown while converting tables to HTML, math to LaTeX, and graphics into bounding‑box annotations, and achieves this by integrating a SigLIP2‑based vision encoder with the Sarashina2.2‑3B‑Instruct LLM for high‑resolution vision‑language understanding.
Resources in similar research fields
Pix2Textbreezedeus3.3kAn Open-Source Python3 tool with SMALL models for recognizing layouts, tables, math formulas (LaTeX), and text in images, converting them into Markdown format. A free alternative to Mathpix, empowering seamless conversion of visual content into text-based representations. 80+ languages are supported.MultilingualShared: OCR & document images・Vision & multimodal
mokurokha-white1.8kRead Japanese manga inside browser with selectable text.GitHubShared: OCR & document images・Vision & multimodal
yomitokukotaro-kinoshita1.7kYomitoku is an AI-powered document image analysis package designed specifically for the Japanese language.GitHubShared: OCR & document images・Vision & multimodal
meikipoprtr46734universal japanese ocr popup dictionary for windows, linux and macosGitHubShared: OCR & document images・Vision & multimodal
Poricomblueaxis447Optical character recognition in manga images. Manga OCR desktop applicationGitHubShared: OCR & document images・Vision & multimodal
owocraurorawright317Optical character recognition for Japanese textGitHubShared: OCR & document images・Vision & multimodal