Generating Manga from Illustrations via Mimicking Manga Creation Workflow - A Framework for Layered Outputs

Aug 28, 2026

Uploaded by Oliver

Generating manga from digital illustrations uses a professional workflow with three indispensable steps: line drawing for structure, regular screentones for shading and occlusion, and irregular screen textures for background layouts and special effects. This work proposes a data-driven deep learning framework that converts an illustration into three corresponding components—manga line drawing, screentone layers, and irregular texture masks—enabling direct composition and further retouching. A large human-annotated dataset supports human-in-the-loop creation, and user studies together with qualitative evaluation confirm the practical usability of the generated layers for manga artists.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages10

Reading Time25 min

AccessFree to download

Introduction

Problem and motivation

Observed manga creation workflow

Proposed deep learning framework

Generating Manga from Illustrations via Mimicking Manga Creation Work􀀃ow Lvmin Zhang Soochow University / Style2Paints Research China [lvminzhang@acm.org](lvminzhang@acm.org) Xinrui Wang The University of Tokyo / Tencent China secret [wang@outlook.com](wang@outlook.com) Qingnan Fan Stanford University United States [fqnchina@gmail.com](fqnchina@gmail.com) Yi Ji Soochow University China [jiyi@suda.edu.cn](jiyi@suda.edu.cn) Chunping Liu Soochow University China [cpliu@suda.edu.cn](cpliu@suda.edu.cn) Abstract We present a framework to generate manga from digital illustrations. In professional mange studios, the manga create work􀀃ow consists of three key steps: (1) Artists use line drawings to delineate the structural outlines in manga storyboards. (2) Artists apply several types of regular screentones to render the shading, occlusion, and object materials. (3) Artists selectively paste irregular screen textures onto the canvas to achieve various background layouts or special effects. Motivated by this work􀀃ow, we propose a data-driven framework to convert a digital illustration into three corresponding components: manga line drawing, regular screentone, and irregular screen texture. These components can be directly composed into manga images and can be further retouched for more plentiful manga creations. To this end, we create a large-scale dataset with these three components annotated by artists in a human-in-the-loop manner. We conduct both perceptual user study and qualitative evaluation of the generated manga, and observe that our generated image layers for these three components are practically usable in the daily works of manga artists. We provide 60 qualitative results and 15 additional comparisons in the supplementary material. We will make our presented manga dataset publicly available to assist related applications. 1. Introduction Generating manga from illustrations (Fig. 1-left) is an important task in high demand. The expansion of manga market and the rarity of manga artist have caused many manga companies to recruit a large number of digital painting or illustration artists and train them as manga creators. Training an artist to master the unique manga work􀀃ow, e.g., inking strategy, screentone management, texture applying, Figure 1 . Our framework automatically generate the right highquality manga given the left illustration. Zoom in to see details. etc., is 􀀂nancially expensive and can often take weeks or even months. To reduce such training costs and speed up the producing, many manga companies have began to adopt techniques that generate manga from generic art forms like illustrations and digital paintings, so that the costs to train artists can be saved, and those newly hired digital illustration artists can be free from learning extra skills and can create manga directly in their familiar digital illustration working environments. The software Clip Studio Paints (CSP) [6] and Adobe After Effects (AE) [1] are typical examples with many plugins and online tutorials [10, 11, 7, 9, 8] for editing illustrations to obtain manga manually. The widespread popularity of those tutorials and plugins veri􀀂es the signi􀀂cance of the problem to generate manga from illustrations. This paper starts with a key observation: the unique visual appearance of manga comes from the unique manga creation work􀀃ow. As shown in Fig. 2, we verify this observation Figure 2 . Justi􀀂cation for our motivation: the classic professional manga creation work􀀃ow. Artists 􀀂rst ink structural line drawings, and then apply regular screentones and irregular screen texture to obtain the 􀀂nal manga. by studying the manga creation work􀀃ow in professional studios. Firstly, artists draw line drawings as the initial outlines in manga storyboards, e.g., the artist delineates the line structure of the girl portrait in Fig. 2-(a) . Secondly, artists paste screentone sheets with different regular patterns onto the region

FAQs

What three components does the framework generate from an input illustration?

It generates manga line drawings, regular screentone layers (as segmented classes), and irregular screen textures via a predicted texture mask.

Why is mimicking the professional manga creation workflow central to the method?

The paper argues that manga’s distinctive visual appearance comes from the specific studio workflow, where inking lines, applying screentones, and pasting irregular textures each contribute essential visual qualities.

How is the dataset created and what role does it play?

Artists annotate a large-scale dataset in a human-in-the-loop manner, providing illustration-to-component supervision used to train a hierarchical neural network for step-by-step conversion.

Aug 28, 2026

Uploaded by Oliver