00:00The world of Japanese manga is experiencing an unprecedented boom outside Japan.
00:04However, only a tiny fraction of the works are translated into other languages.
00:08Today, only about 14,000 volumes, less than 2% of the total published, are available in English.
00:15The main challenge is that conventional translation and localization requires many steps.
00:19Translation, revision, lettering, graphic editing, and covers thousands of pages.
00:24Publishers lament that this process is massive and costly,
00:27translating a single volume can cost around $7,000 and take months.
00:32Imagine a world where every Monda story, whether action or fantasy,
00:36is available to readers around the globe almost as soon as it is released in Japan.
00:40Could artificial intelligence, AI, make this possible in a faster, cheaper, and more accessible way?
00:46In this documentary, we explore how new AI technologies are transforming Monda translation,
00:52who is investing in them, and what translators and publishers say about the future that lies ahead.
00:56Several startups and major Japanese publishers have placed their bets on machine translation.
01:02For example, Orange Incorporated, a Tokyo- and U.S.-based venture,
01:06has raised nearly 2.92 billion yen, $19.4 million, to create AI-powered Monda localization tools.
01:13According to the company, its system, called Factory,
01:16uses computer vision and neural networks to process Monda pages and speed up translation.
01:21With this investment, they hope to increase the current production rate fivefold and reach about
01:26500 volumes per month translated into English. In other words, Orange claims that its technology
01:31would increase the amount of Monda translated tenfold. Meanwhile, the startup Mantra, which emerged from
01:36the University of Tokyo, was founded in Tokyo, specializing in AI for entertainment. In 2024,
01:43media outlets such as EFE and Nikkei reported that Mantra had already attracted around 78 billion yen,
01:48Euro 4.6 million, from major publishing giants, Shuisha, Shagakuken, Katakawa, Square Enix, among others.
01:56Its platform automatically analyzes the panels. First, it detects the text of the dialogues with
02:02image recognition, translates it using AI, and then allows a human to review and correct the result.
02:07According to Mantra, this system can cut the time it takes for traditional manual translation in half.
02:12They also claim that the AI error rate in English is very low, only 1.6%, and is always supervised by a
02:20human translator. The Japanese publishing industry has welcomed these advances as a way to expand its
02:25markets. The Japanese government, for example, has set a goal of quadrupling exports of cultural content,
02:31manga, anime, video games, by 2033, and sees machine translation as a means of exporting more comics abroad.
02:38As Orange states in its public presentation, its mission is to create a world where everyone can
02:43enjoy manga by making content available globally in an affordable way, similar to what Spotify did
02:48with music and Netflix did with movies. Orange Inc. announced $19.4 million in funding to increase
02:55the speed of manga localization fivefold. Mantra, with heavy investment from Japanese publishers,
03:00claims that AI can cut conventional translation time in half. Today, only 2% of Japanese manga is
03:06translated into English, so the industry is looking to change those numbers.
03:10Orange Inc. US, Japan developed the AI localization tool factory and is preparing the Amaki reader
03:16platform. In 2024, it managed to create official simultaneous translations on social media for indie
03:22manga thanks to its service. For example, the comic Niko Aji was published in Japanese and English
03:28simultaneously on Twitter for months. Orange also collaborated with Shuisha, Manga Plus,
03:33on AI lettering, although its first samples were criticized for their visual quality.
03:38Mantra, Japan its platform aims to serve all the major publishers that have invested in it.
03:42It is already said that Shuisha used Mantra to translate One Piece and Spy X Family into Vietnamese.
03:48Other Japanese companies are expected to customize this AI to expand their translated catalog.
03:53Japanese publishers, Shuisha, Shagakuken, Katakawa, all have invested in these technologies.
03:59Shagakuken even launched a manga reading app in the US that publishes works translated with Mantra.
04:04Kodansha and other major publishers, not mentioned in the news, are also studying AI.
04:09In general, publishers are looking to accelerate the simultaneous publication,
04:13Simulpub, of manga in multiple languages and combat digital piracy by offering official
04:19translations more quickly. Global distributors such as Crunchyroll, mainly anime,
04:23have tested AI for subtitles. In the fan scene, there are already AI-based tools, GPUs, OCR,
04:30browser plugins that allow raw comics to be translated at home, but they lack official support.
04:35AI promises several advantages over traditional manual translation. Automatic tools work almost
04:40instantly. For example, Orange expects to deliver English versions five times faster,
04:46and Mantra claims to be able to cut the time of a manual process in half. By automating repetitive
04:50tasks, companies report huge savings. Orange even claims to have achieved up to 90% savings in
04:56localization time and cost. Translating a volume, currently costing $7,000, could be much cheaper
05:02with AI. By making it cheaper and faster, thousands of extra titles could be translated. This would give
05:07readers around the world, not just English speakers, access to previously unpublished works.
05:12Orange plans to start with English, US, and then integrate more languages, from Hindi to Spanish.
05:17Mantra already offers translations in 18 languages and aims to be the choice of major publishers.
05:23AI facilitates simulpubs, where the chapter is released in Japan and in other languages on the
05:28same day. This is already being done by official platforms, Mondo Plus, Shonen Jump Alpha, and AI
05:34would further accelerate this synchronization. By translating faster, the incentive for fans to
05:39seek out pirated versions is reduced. According to the industry, making more works legally available
05:44abroad leaders' mass piracy. List of specific examples, Orange plans to reach 500 volumes per
05:50month. Thanks to its AI, Shusha has translated comics into rare languages, Vietnamese, ahead of
05:56schedule. The Japanese government believes this will help quadruple the global revenue of its
06:00entertainment industry by 2033. Despite commercial optimism, several sectors warn of problems.
06:06Translation quality, languages such as Japanese have cultural nuances, puns, and unique expressions that AI
06:12may not grasp. Professionals have complained that sloppy translations damaged the reputation of
06:17Japanese Mondo abroad. The Japanese Translators Association NAVCA warned, reckless machine translations
06:23are directly linked to a decline in the reputation of Japanese content abroad. Technical aspects,
06:28Orange's results in Mondo Plus, Rugby Rumble chapter, were criticized for gigantic and poorly adjusted
06:34fonts. Although Orange announced a 90% savings, the reality was a final quality much worse than human,
06:40according to professionals. Several letterers tweeted indignantly that the result doesn't look
06:44professional. We do not quote tweets directly, but Animononikas reported that several experts
06:49considered the visual finish unacceptable. Employment of professionals, translators, editors,
06:54and letterers fear losing their jobs. A trade union asked the government to regulate the use of AI
06:59in anime and manga translation. They argue that replacing translators with machines may violate moral
07:05copyright, control of the original work, and make salaries more precarious. According to a CISAC study,
07:11soon more than half of the revenue from audiovisual adaptations will be generated by AI. Errors and human
07:16supervision even advocates acknowledge that AI is not infallible. Mantra, for example, emphasizes that
07:22its AI always goes through human review. In practice, AI can translate the bulk of the text, but then a
07:28professional translator is needed to polish idioms, context, and correct serious errors. Language and variety,
07:34translating a book or script is easier than a manga, because the latter has graphic considerations,
07:39space and speech bubbles, and sound effects. AI must also relabel texts and even redraw onomatopoeia,
07:46which is complex. For now, most systems combine AI plus manual work, which is why they talk about AI
07:52assisted. Reader reaction, manga fans noticed when manga plus used AI or AI tools in a simulpub. Many
07:58demanded a return to human translators because of the poor finish. The annoyance stems from the fact that AI
08:03does not feel the story, and its mistakes are very noticeable when consuming the works.
08:08Experts agree that AI is a tool, not a total replacement. A professional translator consulted
08:13indicates that the biggest advantage today would be for high-volume work, thousands of low-priority
08:18momda, while for important licenses, investment in human translators will continue. Others believe
08:24that in the long term, AI will set the standard in the mass market, and only special or cult volumes
08:29will remain as manual craftsmanship. The publishing and technology industries see a mixed outlook.
08:34Expanded global access, if everything goes according to plan, many previously unpublished
08:39manga, from classics to new releases, will be available in multiple languages within months or
08:44weeks of release. Apps such as Amaki, from Orange, and other international stores aim to bring together
08:50automatically translated titles alongside content curated by human editors. AI specialization,
08:56companies like Mantra will continue to train their models in manga jargon, samurai terms,
09:01technicalities. Over time, translation quality will likely improve, reducing the need for
09:06corrections. Regulation and standards, given the concerns of unions and authors, we may see rules
09:11on when and how to use AI. Japan and other countries are mindful of the balance between innovation and
09:16creator protection. Role change, translators may focus more on revision and cultural localization,
09:22leading the mechanical parts to machines. Editors and localizers will still be needed to ensure
09:27overall consistency. Market scalability, analysts believe that the global manga market, currently
09:32worth several billion dollars a year, will grow if it breaks down language barriers. Currently, 66% of
09:38manga in Japan is digital, but outside Japan, it is only around 20%. Multilingual platforms such as
09:45Netflix for manga could bridge that digital divide, driven by machine translation. AI manga translation
09:51is already a developing reality, startups and renowned publishers are investing heavily to speed up the
09:56process. On paper, AI offers faster and cheaper translations, making it possible to distribute thousands
10:02of titles that currently never reach international audiences. However, experts warn that quality and
10:07accuracy remain a challenge. As the case of Orange and Rugby Rumble shows, the end result leaves much to be
10:13desired despite the announced savings. In this documentary, we have seen that, for now, AI works
10:18better as an assistant than as a complete replacement. It analyzes images, suggests translations, and speeds
10:24up the work, but it still needs careful human supervision. Ultimately, AI could make manga translation faster,
10:31cheaper, and more widespread, but not without risks. Current data indicates that more manga is being
10:36translated than ever before, and the trend is on the rise thanks to these technologies. The balance between
10:41innovation and quality will determine whether readers around the world will soon enjoy an explosion of
10:46content, or whether the Manda community will impose limits to preserve the original essence. Only time,
10:52and the voice of translators and readers, will tell how much of global Manda will end up in our hands
10:56thanks to AI.
10:57Thanks to AI.
Comments