FuguTabetai July 2009
FuguTabetai is a translator, a programmer, and the founder of MangaTranslation. FuguTabetai has been involved in the scanlation scene since the early 2000s, and is the creator of Great Manga Application Onidzuka, a program made to facilitate the scanslation process in the early days of scanlation.
FuguTabetai: I'm FuguTabetai, just your average every day translating fish. If you are determined (or even not) and have a few minutes, you can probably find out who I am in real life, but the pertinent bit is that I've been translating manga since about 1999. In October of 2001, I registered MangaTranslation.com and started putting stuff up on the web.
FuguTabetai: I started translating manga in 1999. At the time, I was in Japan for the summer doing an internship at a company there. I had studied Japanese for a while, maybe three years or so, and I wanted a way to improve my Japanese language through daily study in a fun way. So I started to read manga. One of the books that I liked was Great Teacher Onidzuka. When reading manga just for fun, it is very easy to go quickly and gloss over words or expressions that you might not completely understand. I decided to translate from Japanese to English so that I would be forced to understand the Japanese.
I have a background in Computer Science, and in particular natural language processing, so I thought it would be fun to make data that would be potentially useful for machine translation research. So I decided to write out the Japanese and my English translation in a way that would be computationally accessible.
FuguTabetai: For those of you that might not know, Great Manga Application Onidzuka was written for two reasons:
- To keep track of my translations, and align the Japanese and English translation data in a useful way.
- I was interested in learning about Java Swing interfaces and XML processing in Java, so I thought I would write a real application to learn about it.
GMAO is an application that basically loads up a picture and lets you draw different kinds of "bubbles" on top of the picture. You then can enter text into the bubble and the software will automatically fit the text into the bubble. It doesn't really do the best job in the world of that, but it works well enough for my purposes. You can then save the picture in a couple of different formats. The way I use GMAO is to draw a bubble over the dialogue in the manga, then enter the Japanese text, and the English text into the bubble. You can choose which language to display. When I'm done with a page, I click the export button, and save out a JPG of the English translation.
I wrote the program in Java, and the data is saved as XML. This is important because XML is a computer-accessible format, so at some point in the future I could potentially use the data to train a machine translation system. I actually did do that a summer ago, but the results were not all that great; you really need a lot more data than just ~40 volumes of manga to learn a translation model. I am interested in using available newspaper corpora to build a language model, and then use the manga data on top of that, but I have not had time to continue with that project.
The program itself was written mostly using Eclipse, and I learned a lot about Java. Now, almost ten years later, the main thing I learned is that the program itself really sucks. I need to go back and write the thing over from the ground-up. It was a great way to teach myself Java and Swing though.
Another fun thing I did with the system was to write it as a client-server model so I could set up a GMAOServer on some machine, and then connect to it from any machine on the Internet with Java and translate without loading up the scans or data on the client machine: that was all downloaded over the Internet. That was fun, but I haven't been using that functionality for a while.
As for the reception of GMAO, it was better than I expected. I think currently there are about two or three people in the world using GMAO (including myself). That is one or two people more than I expected.
FuguTabetai: The plan with GTO was to have other people use GMAO to translate it into their language, so we could have an English translation by me (and perhaps other people), maybe a translation in Chinese, into Spanish, German, whatever. You can choose what language to export in GMAO, so you only have to do the work of adding the bubbles once, then it will automatically do the text layout and export.
That did not happen at all. I did have some help translating into English, but I didn't usually like the translation since it was from the Chinese published version to English, and it was different from the Japanese version. Just slightly.
When I was translating GTO I also used some of the features of GMAO to make the bubbles different colors based on the character that was speaking. This looked horrible. Nobody liked it. I wasn't really interested in creating a "scanlation," I wanted to make translations mostly for myself. But in the end people were complaining about how bad things looked. I decided that I wanted to try to make a more traditional "scanlation" where things looked nice.
So I started to translate Tenjo Tenge, a manga that I picked up because the cover looked interesting, and the first volume was fun. I made it look like a normal scanlation project. I think it turned out pretty well. Of course, it still never looked as good as a project where a human does the text layout and cleanup, but it works well enough.
FuguTabetai: Technology and software is completely irrelevant to scanlation. Most groups do things based on pure manpower: someone scans, someone cleans the scans, someone removes the text, someone translates, someone re-inserts the text, and someone does quality control. Usually the process looks like that. All of this can be done with ten year old versions of Photoshop, or for free with the open-source program The Gimp. As far as I can tell, everybody just uses a pirated version of Photoshop and text files.
I was interested in doing something for myself. I think that it is a shame that in the traditional translation project you lose the valuable Japanese to English translation data, and in the end have a picture with English text on it that is not accessible to computers. You can't search the data at all. So I wanted to make something that does what I want to do, and I wrote my own tools.
The only interesting thing that I saw was Mr. Dummy's translation project which is an online system for translating manga. You put a box around the bubble, and enter your translation. Then it will pop up the translation when you mouse over the bubble. GMAO can also export data in the form, by the way. It can also export a HTML translation script.
The one thing that I really did like as far as technology goes was the introduction of BitTorrent. I was one of the first people to use BitTorrent—I think I started with the Python v0.2 tracker—and early on even wrote a small patch for it. I saw the announcement of its release on Slashdot and was a very early adopter. Anyway, I started using this new "tracker" and "torrent" technology to distribute chapters of my translations as I finished them. I am still using the Python based tracker today because I never outgrew the functionality that it offered. Now using BitTorrent for distribution is a common practice.
FuguTabetai: MangaTranslation is a group headed by a translating fish. It has been around since 1999, but the website has only been around since 2001. The history is as above. I usually translate stuff that I like; I try to translate a bit each day, but lately haven't had the time for it.
FuguTabetai: I have never been in the scanlation community really. I do my own thing in a way that is more or less different from what other groups do. I am not even sure how many translation projects were around when I started in 1999, but I don't think there were that many. The scene really seemed to explode in 2002 or so. MangaTranslation wasn't really received by the community I don't think because I didn't really do any sort of advertisement or anything.
FuguTabetai: I thought Manga Jouhou did a great job of creating a community and getting the word out on projects that existed. I really liked the part of their forum where translators gathered and could ask each other questions, that was very useful for me. After Manga Jouhou, eventually a bunch of other sites sprung up and started doing RSS feeds and the like of new releases, which really increased the number of people out there downloading stuff.
FuguTabetai: The biggest roadblock has been finding time for translation. I did once get contacted by TokyoPop about GTO, and then I quit translation. They had asked about possibly using my translation, which I was fine with, and I sent it on to them. In the end I don't think they used my translation, but I never really looked into it. Perhaps some translator there had it as reference for the first few volumes. Who knows?
FuguTabetai: A typical translation process looks like this:
- Scan something.
- Run the scans through a Gimp script that resizes them, reduces the color space, adjusts levels, and saves a version for translating from. This is the cleaning step.
- Translate with GMAO.
MangaTranslation.com is a single-fish operation usually, although I have had some great help from Ki Shodar with Tenjo Tenge over the years.
FuguTabetai: Tenjo Tenge is probably the most popular project. I don't know that we have had any memorable stories. You can find some chapters translated by Evil Genius because I'm too slow. Since I mostly translate for myself though, I just keep going at my slow pace.
FuguTabetai: I don't really think that is important. People should do what they want to do. In the end everyone in the fan translation community is committing copyright infringement, so having rules within the community about who should be able to break the law using which series is somewhat ridiculous.
FuguTabetai: I haven't followed any translation projects; any time I have for manga I would rather spend translating rather than reading.
FuguTabetai: I think that using manga to study the Japanese language is a great way to keep language study fun. One of the goals I have for GMAO is to be able to use it to produce material that is useful for language learners. I have always been surprised at how few people are interested in this though. Still, there are some, so that is better than none. I usually get perhaps 5 emails a year about manga and Japanese language learning. I get closer to 40 from people just asking me to translate something for them.