alternative title: awesome-fansubbing
This isn’t actually a guide, just a bunch of links to other peoples’ guides for specific things. If you follow them, you should end up with some at least semi-decent subs.
It covers most everything except the actual translation of the content. Mostly oriented towards anime, but should apply to most things (with some notable exceptions).
explanations of the different roles are taken from guide.encode.moe, and are licensed under CC-BY-SA 4.0. Some portions are omitted, but otherwise unchanged. Aegisub Configuration is nicked from the GJM fansubbing page.
For more resources, and to talk with actual people, join the GJM discord server and give yourself the fansubber role. I do not recommend participating in the general discussion channels.
The new fansubbing wiki also has an equivalent page, which ideally will replace this page eventually.
this will all be gone once 3.4.0 comes out (hopefully)
The current Cartel Approved™ build is arch1t3cht’s fork. It has lots of new shiny features: subtitle folding; new video providers (and the return of an old one); performance improvements; hidpi support; stereo on linux (!); video panning; a visual perspective tool; new lua stuff; and lots of bugfixes.
It runs on all three major operating systems, but you’ll have to build it yourself on Linux.
The newest shiniest features (folds, new lua stuff) won’t work on other builds.
If this doesn’t sound appealing to you, read on.
don’t use mac; no clue
rcombs-local branch. Clone recursively, like in the windows instructions.Doesn’t really matter, most of the forks are just to keep dependencies up-to-date on windows.
If you use 3.2.2, make sure to configure it correctly.
Upon starting up Aegisub the first time, you should change the following settings:
View > Options > Advanced > Video
libass.Restart Aegisub for the changes to fully take effect. Note that script settings may override the colour space.
See guide.encode.moe for why these settings are important.
[13:20]arch1t3cht: you need to install ffms2 from git, not from whatever your distro has
[13:20]arch1t3cht: for example ffms2-git from the aur if you're on archBut if it ain’t broke, don’t fix it. I used distro ffms for years and I never had any problems even with the things people said were broken.
Alternatively, you can use a different video source altogether. arch1t3cht’s fork has a few, apparently everything should work out of the box now as long as you have python and vapoursynth installed.
No. Read this.
(you could probably get away with it for some roles, but it’ll be a significantly worsened experience.)
Sure, has some cool features, and is definitely better than the other not-aegisub options out there.
Be aware, however, that it has issues with automation scripts. Be prepared to work things out on your own if stuff breaks.
Actually, I think all you need to do is upgrade the bundled moonscript, and then you’ll be okay. Note that I have not tested this in the slightest.
Learning the source language is outside the scope of this guide.
TED translations: How to Tackle a Translation
TED translations: The translator’s research toolbox
Translation theory overview [PDF]
An Open Letter on Translating (Martin Luther, 1530)
In Other Words - A Coursebook on Translation - ebook ISBN 978-1-3156-1918-7
The Editor is responsible for making sure that the script reads well. Depending on the source of the script, this may mean grammatical corrections and some rewording to address recommendations from the Translation Checker. However, more often than not, the job will entail rewriting, rewording, and characterizing large portions of the script.
The Translation section is worth a read. Yes, really.
Collectr’s Curmudgeonly Guide to Editing
Light/Kaleido’s style guide [wayback machine]
sci.lang.japan FAQ: What are these pseudo-English words like salaryman? (wasei-eigo)
sci.lang.japan FAQ: What “false friends” are there between Japanese and English?
Wikipedia: List of gairaigo and wasei-eigo terms
editing style comparison meme [large (high-res, ~5MB) image]
read books
How to break lines (adding \N, not destruction of the script).
fwiw, UA’s linebreaker script seems to do what I would’ve done anyway ~70% of the time. Make sure to check its work so you don’t miss the other ~30%.
The Timer is responsible for when the text representing spoken dialogue shows up on screen. The timing of subtitles is much more important than one might assume. The entrance and exit times of the subtitles, or a fluid transition from one line to the next, can make a large impact on the “watchability” of the episode as a whole.
Read WhyNot’s guide [PDF] for the basics. Feel free to ignore what it says about settings.
Then read unanimated’s timing guide (every page) for the slightly-less-basics and better settings.Use the spectrogram, not the waveform. It might take a bit of getting used to, but it’s worth it.
Alternatively, use PhosCity’s Timing Assistant script. It’s like the TPP, but it only runs one line at a time, and you can check and change its decisions as it makes them. This means that, in theory, you can get the entire thing done in one pass. You should still do a QC pass though, if only so you can watch the episode.
You should generate your own keyframes, the ones in the video probably won’t be very accurate, or just straight up wrong. This especially applies if you’re using untouched WEB-DLs or TV caps.
There’s a nice batch script that comes with all the stuff you need. You should be able to just drag your video onto it, but I haven’t personally tested it. Apparently you can also add it to the Send To menu.
get ffmpeg
get scxvid-standalone - you’ll need to compile it on not-windows, but this only took a second or two even on my ancient hardware.
run ffmpeg -i vid.mkv -f yuv4mpegpipe -vf scale=640:360 -pix_fmt yuv420p -vsync drop - | scxvid vid_keyframes.log
If it misses frames, try removing the -vf scale=640:360 bit. It’ll take longer, but it’s much more accurate.
there’s also myaa’s keyframe script. WWXD, the thing it uses to actually generate the keyframes, claims to be ~6× faster than scxvid.
You’ll need a working VapourSynth install to use it though. Setting that up is outside the scope of this page.
If you want more control over your keyframes, use this script by Setsugen no ao. You’ll need working vapoursynth etc for this too.
Zahuczky’s Karaoke Timing guide. The 0th part of a larger KFX guide.
WhyNot’s guide [PDF]
don’t.
but if you must, see WhyNot’s guide (again) for an idea of what it is and how it works. then forget everything it says and use my karaoke2alpha script instead. K-Time the line, with the words you want to appear as syllables in the karaoke, then run the script. That’s not a great explanation, but have a play around with it and it should become relatively clear.
Typesetters (abbreviated TS) are responsible for the visual presentation of translated text on-screen. These are generally called signs. Almost every sign the Typesetter works on will be unique, requiring ingenuity, a wild imagination, a sense of style, and a high degree of attention to detail. The Typesetter’s goal is to produce something that integrates so well into the video that the viewer does not realize that it is actually part of the subtitles. Something to remember about typesetting is that there is no one way to typeset a sign. There are, however, incorrect ways that are not visually pleasing, do not match the original well, are difficult to read, or are too heavy (meaning computer resource intensive).
I don’t typeset, so this section isn’t nearly as good as it probably could be. Contributions welcome.
There’s only one really good piece of documentation. Everything else is either word-of-mouth, or you have to figure it out on your own. Yeah, it sucks. I once again encourage joining the GJM discord server and getting advice from real human beings.
The new fansubbing wiki is attempting to change this by documenting things. It’s not gotten very far yet, but it’s on the right path, I think.
unanimated’s typesetting guide [unreadable colour scheme]
>vertical text [youtube]
AI2ASS - A script to export ASS vector objects from Adobe Illustrator. Very powerful, very addictive.
Typesetting in Adobe Illustrator
Masking with Photoshop and Illustrator
svg2ass - Alternative to ai2ass. much worse, but works for anything that can output svg.
Aegisub-Motion - The motion tracking script.
Aegisub-Perspective-Motion - A recent innovation that tracks perspective as well.
Motion tracking with Mocha and Aegisub-Motion
UA’s mocha guide [outdated?]
Tracking things in blender [dodgy english]
There are lots of fonts that the studios use pinned in #typesetting in GJM. Not linked here so I don’t anger the hosting provider.
You don’t need to find the exact font the studio used. Even if you do, the latin letters might not be nearly as stylised as the kanji. As long as the font matches decently well, no one will care.
Or, alternatively, do something completely different. As long as it makes some sense for it to be there, you’re probably ok.
At some point, it’ll start making sense to use a font manager. I hear FontBase is good.
Very important. So important, in fact, that I’ve split it off from the typesetting section and into its own.
Don’t skip this bit, even if all your other typesetting is just \an8.
Underwater’s Styling Guide Rant [wayback machine]
Subtitle Appearance Analysis Part 1: The Font
Subtitle Appearance Analysis Part 2: Font Size
Subtitle is important - Austin Powers [youtube]
Remember,
TL;DR: Just steal GJM-Main and have done. (style lines for 1920x1080. font here)
For an example of what not to do, go to the cat site and search for anime chap.
Like typesetting, but this time there’s barely even one piece of documentation. Again, GJM discord.
You’ll probably need to know how to typeset so you can express what you want to the templater. You’ll also need some K-Timed lines, obviously.
Zahuczky’s KFX Guide [incomplete, work in progress(?)]
Jockotan’s blog [incomplete, dead]
A Programmer’s Guide to Karaoke Templaters
Quality Checkers (abbreviated QC) are often the last eyes on an episode before it is released. They are responsible for ensuring that the overall quality of the release is up to par with the group’s standards. They are also expected to be familiar with the workflow and many intricacies of every other role.
Collectr’s Curmudgeonly Guide to QC
Light’s QC Ramblings (from MRF #general)
The Dunning-Kruger effect (why QC is important (among other reasons))
You can only really QC well if you know what you’re looking for - you basically need to know all the other roles. This makes it probably one of the hardest roles to do well.
go read all the other categories
This section is not very good.
.aegisub folder. (read: never)Have a look through the DependencyControl script browser and the fansubbing wiki’s list of scripts not available in DependencyControl.
There used to be a bunch of links here. I’ve commented them out, but you can still see them if you read the page source.
completely optional, but writing your own scripts for repetitive tasks can save you a lot of time in the long run (and is fun for nerds like me).
read lyger’s guide, skim unanimated’s extra stuff, RTFM, and google any lua stuff you don’t know.
Try not to blindly copy+paste from UA’s scripts if at all possible. The code is… of poor quality, to say the least.
Prior knowledge of programming will be very useful. A decent grasp of maths would definitely help if you write typesetting scripts.
You can also use MoonScript, but figuring that out is left as an exercise for the reader (read: i don’t know how, and i dont like it anyway).
Try not to reinvent the wheel too much.
I do encourage wheel reinvention to some extent. It’s a good learning experience, and you do (hopefully) get something useful out of it at the end. Also, making your own tools is fun and satisfying.
If, however, you value your time, quit fansubbing and go do something else. make sure someone else hasn’t already done what you want. If that someone else is UA, please, go ahead and do it anyway.
I don’t know whether my use of the guide.encode.moe descriptions qualifies as fair use/fair dealing, so: This “work” is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.
