Mantra, an AI Manga Translation Company, Jointly Develops Manga Typesetting Specification Engine with Square Enix

— Streamlining approximately 3,000 hours of annual manual labor for editors; phased rollout begins following successful beta testing —

Mantra Inc. (Headquarters: Bunkyo-ku, Tokyo; CEO: Shonosuke Ishiwatari; hereinafter "Mantra"), a developer of AI technology specialized for manga, has jointly developed an AI technology with Square Enix Co., Ltd. (Headquarters: Shinjuku-ku, Tokyo; President and Representative Director: Takashi Kiryu; hereinafter "Square Enix") to automate the "typesetting specification" (shokusetsu shitei) process in manga editing. Following the completion of beta testing, phased implementation has begun within Square Enix.

Note: This technology is designed to assist in the arrangement of text and layout for existing manga manuscripts. It does not generate, alter, or train on the actual artwork or creative expression of the manga.

This project was realized through a collaboration established following Square Enix Holdings' investment in Mantra in 2024. It applies the image recognition and style estimation technologies that Mantra cultivated through the development of its manga translation tool, "Mantra Engine," to the Japanese manga editing process. Moving forward, the two companies will continue to verify and improve the system within Square Enix to enhance usability and operational efficiency, with the long-term goal of offering the product to the entire manga industry.

■ What is "Typesetting Specification"? — Why it takes 3,000 hours a year

"Typesetting specification" is the process of individually designating the font type, size, style, and placement for every line of dialogue in a manga. For instance, even for the same character, different fonts and styles are used for daily conversation, shouting, internal monologues, and narration.

Applying the appropriate font is a crucial step that directly impacts the storytelling and the reader's immersion. Editors currently perform this manually for every single line of dialogue—ranging from dozens to hundreds per chapter. At Square Enix's editorial departments, more than 3,000 hours per year are spent on this task. As the volume of manga publications continues to grow, streamlining this process has become an urgent challenge for the editorial field.

Manual typesetting specification with a dedicated scale.
 ⓒMiyuki Tonogaya/SQUARE ENIX

■ Overview of Joint Development

In this joint development, Mantra applied its manga image analysis and style estimation technologies to create an "AI Typesetting Specification Tool." Key features include:

  1. Structural Analysis of Speech Bubbles and Dialogue: Automatically analyzes the shape of bubbles, text content, character count, and number of lines.
  2. Automatic Text Layout Estimation: Proposes the appropriate "kyusu" (character size) and positioning based on the size of the bubble. It also estimates the suitable font type and style by identifying the bubble's shape (e.g., standard, thought, scream, or narration).
  3. Custom Design for Editorial Needs: Based on interviews with Square Enix editors, the tool includes an intuitive UI and a "Typesetting Preview" function to ensure practical usability for professionals.
AI Typesetting Specification Interface
The AI analyzes the "speech bubble shapes" and "character data as symbols" from the manuscript on the left, then proposes font styles and character sizes within the bubbles of the manuscript on the right.
ⓒMiyuki Tonogaya/SQUARE ENIX
■ Beta Test Results and Feedback

A beta test conducted within Square Enix editorial departments (covering 1,516 pages) yielded a 73% overall satisfaction rate. Notably, 100% of respondents expressed a positive stance toward continued use. Regarding operability, 60% found it "easy to use," praising the intuitive UI even for first-time users.

While there is room for improvement in AI analysis accuracy and processing speed—which will be the focus of future development—participating editors commented:

  • "I can now progress with typesetting specifications while handling other tasks."
  • "It saved me the trouble of handwriting font instructions, making the work much easier."
■ Future Outlook

With the beta phase complete, phased implementation is underway. Mantra and Square Enix will continue to refine the tool for better usability and eventually aim to commercialize the product for other publishers and companies.

■ Executive Comments

Katsuyoshi Matsuura, Executive Officer & Publisher, Square Enix Co., Ltd.:

"This tool was developed to optimize the workload of editors by merging the ideas of those on the front lines with Mantra’s technology and the AI expertise Square Enix has cultivated over a decade in digital entertainment. Our goal is not just efficiency for its own sake, but to create an environment where editors can focus on their primary mission: supporting creators' activities, while AI handles routine tasks."

Shonosuke Ishiwatari, CEO, Mantra Inc.:

"Since our founding, Mantra has developed AI to streamline manga translation. I am truly delighted that the core technologies we built for that purpose can now serve the editors who bring manga to life. I would like to thank everyone at Square Enix for their partnership. Mantra will continue to build technology that helps more incredible stories reach more fans."

■ About the Companies

Mantra Inc. Founded in 2020 with the slogan "Manga for everyone, in every language," Mantra develops AI specialized for manga. Its products include "Mantra Engine" and the English learning app "Langaku." The company has received numerous accolades, including the HONGO AI Award (2019) and Google Play Best of 2025 (Self-Growth Category).

Square Enix Co., Ltd. Square Enix is a leading global entertainment company that develops, publishes, and distributes high-quality content including games (HD, online, mobile), manga, and magazines. It also leverages its IP for merchandising, animation, and live-action projects.

Media Inquiries: Mantra Inc.
PR: Funakoshi
Email: info@mantra.co.jp

← Back to all news