Hacker News Reader

Comment by Hamuko | Hacker News Reader

ASS subtitles are particularly popular in anime fansubbing as it allows for advanced typesetting.

You can take it pretty far: https://www.youtube.com/watch?v=U9AgMlHJe7Y

I just watched a fansub recently where they replaced some environmental text in the background, and they used the ASS format to render the english text out of focus and seamlessly match the depth-of-field effect used in the show. Some groups really go above and beyond.

Definitely. For example, MTBB's subtitles for Kimi no Na wa are very impressive.

There's a short showcase of them on YouTube (spoilers): https://www.youtube.com/watch?v=n_c-eKTisI0

It's a shame you basically cannot get anything close to the quality of fansubs from any commercial/legal options.

> It's a shame you basically cannot get anything close to the quality of fansubs from any commercial/legal options.

Fansubbers work for the shows they love while the commercial distributors just pick the cheapest option to get the thing done ASAP, don't matter how sh*tty the result is. There is no chance for quality here.

The same happens with Blurays x rips streams: piracy is the best option (no DRM, no unskippable screens, no ads, no regional restrictions).

Hardware decoders have memory limits and can't render SSA subtitles, which were designed by an insane person and require emulation of quirks of the Win32 font APIs plus a custom buggy implementation of 3D text rendering. (As far as I know, there are only three implementations and I wrote one of them.)

But also, Crunchyroll uses the fansub toolset and SSA subtitles.

The quality of fansubbing varies greatly depending on the translator too. For example, Crunchyroll uses ASS format for their own subtitles (things licensed from other studios seems to use a simpler format), and some translators do a really good job very similar to the fansubbers.

Case in point, I remember watching "Is the Order a Rabbit S3" in Crunchyroll, the English subtitles are pretty basic but the Portuguese (Brazillian) subtitles are great, including things like Karaoke in opening and ending.

> Fansubbers work for the shows they love while the commercial distributors just pick the cheapest option to get the thing done ASAP, don't matter how sh*tty the result is. There is no chance for quality here.

I wouldn’t call subtitles that require about a GB of RAM to process and don’t even get 20fps on an i7-6700 "quality". That’s just a bunch of shitty hacks people accept because there’s no alternative.

ASS is a horrible format (three implementations, only a single well-working implementation, performance hungry, etc), and almost all of the benefits of ASS can also be had with e.g. TTML without the issues.

And stuff like rendering the same line hundreds of times to get a gradient effect or blurred background is absolutely wasteful, especially because it’s just a hack to work around the broken ASS limitations.

Probably because they are forced to use really shitty subbing formats for assisted text in some compliance format :|

Edit: Are those baked into the video or did they just use the ASS format?

They use the ASS format and the text isn't baked into the video. Here's the same video with the softsubs toggled on and off: https://screenshotcomparison.com/comparison/10459

Even the old-school analog EIA-608 captioning system can do positional captions. It's all these modern web players (like Youtube's) that have regressed to a single line of text bottom-center.

YouTube actually supports it through a feature called WebVTT [0], though it is rarely used, an example of it is this video [1] where it is used for placement and lighting up text for karaoke.

The actual code for it is looks like this:

  00:03:31.485 --> 00:03:31.719 align:start position:0% line:0%
  <c.color96D2D3>Hey! nanika ga okoru spe cial </c><c.colorFEFEFE>night</c>
  
  00:03:31.485 --> 00:03:31.719 align:start position:0% line:0%
  <c.color96D2D3>Hey! nanika ga okoru spe cial </c><c.colorFEFEFE>night</c>
  
  00:03:31.719 --> 00:03:32.486 align:start position:0% line:0%
  <c.color96D2D3>Hey! nanika ga okoru special  night</c>
  
  00:03:31.719 --> 00:03:32.486 align:start position:0% line:0%
  <c.color96D2D3>Hey! nanika ga okoru special  night</c>

[0]: https://en.wikipedia.org/wiki/WebVTT [1]: https://www.youtube.com/watch?v=AufydOsiD6M

Just a heads up for anyone who want to see the karaoke, set the CC language to Japanese.

Here's another good example: https://youtu.be/ddWJatRxfz8

And again, set language to Japanese.

Actually most modern systems support image embedding, animations, styling, and positioning with TTML. It’s an official spec, supported by every television, Netflix, Chromecast, etc.

It’s also a format no fansubber ever heard of.

TTML is also supported by pretty much every piece of m pptultimedia equipment out there including chromecast , which is the exact opposite of ASS which is supported by basically nothing.

The fansubbing community is so annoying with its insistence on terrible formats. The community has somehow standardized on 10bit hevc with ASS subtitles. That makes releases so hard to play.

And while there are ways to implement ASS in the browser, those require compiling the one single ASS library that actually works (because it’s an implementation-defined format, ofc) to WASM and running that in the browser. With the expected performance issues.

Now try running that on a Chromecast >_>

The fansubbing community is just so stuck in weird formats which provide no benefit except a virtual moat preventing newcomers.

Can you watch Crunchyroll on a Chromecast? They at least use ASS in the browser.

Sure, but only the english subtitle, and you can’t disable it. For Chromecast, you have to use the burned-in subtitles.

Which is expected, considering the hardware just isn’t powerful enough for ASS subtitles (to get good performance, even on a desktop you’ll want a recent CPU and at least 1-2GB of RAM): https://en.wikipedia.org/wiki/Chromecast#Model_comparison

You can definitely still do some interesting things with YouTube subtitles;

https://www.youtube.com/watch?v=1cqqGOvOGfI https://www.youtube.com/watch?v=QvO8kcWFFT0

I honestly find the idea of the English text inside of the picture as such, blending in perfectly with the scene to be most disorienting.

It suggests that the character itself is writing the English text, or that it magically appears.

If it were simple subtitles, it would be clear that these were translations, and that in the actual story there is only Japanese text.

Notwithstanding the impressive technological nature of it, but that also seems the reason they did so. Methinks it's a case of using cool technology because one can, even though it doesn't produce a more convenient result, however cool it might look.

I cannot agree after watching shows which embedded text well.

Your argument kinda sounds right when you just imagine it... but doesn't hold weight once you actually experience them.

If this text isn't done like this then you have to either omit it entirely, possibly leaving out relevant details, or just put them next to the normal subtitles. The latter doesn't work at all for me, because I can't distinguish which text relates to what quick enough.

It can be overlayed at the correct place, without blending into the picture.

For instance translating the text on a piece of paper by writing the translation above the piece of paper.

How it is written in this case, in charcoal, on the paper, in perspective, seamlessly blending into the paper, makes it seem as if, for whatever reason, the same text was written twice on the paper in English and Japanese.

It's even more unnerving when the text be written and the English pencil charcoal appears out of thin air next to the Japanese charcoal that emerges from a pencil.

Would you know if they are tracking the text frame by frame? That's quite the dedication if they are.

The most popular subtitling software has a script/plugin to export video, which can then be used in a motion tracking program, the data from which can be fed back into the subtitling program.

The few groups that used to do that have stopped making fansubs. Mostly because HorribleSubs took a lot of the market share by just ripping CruncyRoll subs and re-packaging with RAWs.

But now that HS is gone, I'm not sure what's happening to the scene. There is a power vacuum and I'm not sure if anyone took up the space.

I don't want to name them here, but a new group has popped up that does the exactly same thing HorribleSubs did.

Erai-raws has been at it for a while now.

They rip the subs too? I assumed they just ripped the raw video because of their name.

The few anime I've watched from their releases all had subtitles.

There are definitely still anime fansub groups active, although not nearly as many as there used to be.

Mostly they seem to work on shows that haven't been licensed by official simulcast sources, or that are licensed but aren't being released on a timely schedule (ahem Netflix), or projects that for whatever reason they think the official releases haven't done justice to.

Aside from HorribleSubs, these days you can hardly throw a rock without hitting a major streaming service that offers anime. Setting aside whatever free episodes you can get from Crunchyroll et al., Netflix, Amazon, Hulu... the list goes on. And for the majority of people, the services are convenient enough to outweigh whatever video quality advantage that HS or subtitle quality advantage (some) fansub groups offered.

The show I watched was a recent release, but it went through Netflix jail so I guess that's why it got the high-quality fansub treatment. There weren't any official subtitles for groups like HorribleSubs to yoink day of release which bought time for a high-effort group to do their thing.

Subsplease seem to have replaced them.

Isn't that basically crunchyroll taking most of the market by doing sub translations in the first place? Kind of like the netflix / spotify effect on piracy.

Groups like that are a dying breed. Many of them simply fell apart as key individuals quit or drifted away from the scene.

So the translated “fidget” parts are not hardsubbed but actually encoded in a subtitle format and therein animated to move around with the Japanese text?

Quite impressive, notwithstanding the use of bourgeois harem as example material.

Yup. Recently there have been tools that will track selected things for you, but it wasn't too long ago when fansubbers had to manually animate them by annotating keyframes.

That's correct.

> ASS subtitles are particularly popular in anime fansubbing as it allows for advanced typesetting.

Those are just "SSA" subtitles. The difference mattered (ASS was an "Advanced SSA") once but now people just call it "ASS" as a joke. So Google is sort-of right, he could leave it out of the description.

I doubt this guy knows this since it's impossible he implemented SSA on his own due to how complex it is.

Potentially NSFW (heh this is kinda pointless atm).

Also wow, I'm guessing you need to adjust the subtitle with every frame, old-school animation style?

The ASS format has some basic animation with the \move command (moving subtitles using linear interpolation) and the \t command (changing various properties of a line with the possibility of quadric interpolation). This takes care of some cases, but is often insufficient for complex stuff.

Aegisub however has good automation support, and has (had?) a pretty active scripting community. Scripts can do a lot of stuff and often has simple dialog-box UIs. The base scripting support is in Lua, but a lot of scripts are written in a Lua variant called Moonscript. I've never seen it used anywhere else. And, as was mentioned in a sibling comment, there are scripts for importing motion-tracking files into Aegisub.

The ASS format lacks support for a bunch of stuff that people want to do, or the native support is deficient in some manner, so scripting is very important for Aegisub. For example, a line can be given a color gradient by make a one copy of the line for each shade in the gradient (this can be 100+ times). Each copy is then given a different color according to the gradient, and a \clip command is used on each line to only show a 1-2 pixel wide slice of each line. If all those slices are lined up properly and the lines are displayed simultaneously, it gives the impression of being a single line with a color gradient. Tricks like this can actually cause performance problems in rendering the subtitles!

I think this is an interesting little programming niche, but there isn't much complex typesetting being done these days. Tastes have changed, and due to licensed releases there's less need for fansubbing on all but the most niche titles.

The subtitles are basically adjusted every frame but you don't necessarily have to do it by hand. You can for example generate motion data with Blender or Mocha and import that into Aegisub (the software that is used to create the subtitle files).

https://unanimated.github.io/ts/ts-mocha.htm

https://github.com/TypesettingTools/Aegisub-Motion/wiki/Appl...

Before newer tools automated the process by tracking motion of a target, folks wouldn't have to adjust them frame by frame per se but would instead annotate the subtitle script with keyframe times and positions. The video player would know to interpolate between those annotations.

(Shameless plug for an abandoned project.) If you want to try out the ASS format for yourself in the comfort of your web browser: https://arnavion.github.io/libjass/demo/

Commie should be a national treasure

Warning: NSFW for those clicking this

Reply on news.ycombinator.com