SSA, ASS, USF, AS5, SSF - subtitle format wars

[ArchMageZeratuL]

As a few of you might know, Gabest (creator of VSFilter and Media Player Classic) has began working on a new subtitle format, temporarily called "Structured Subtitle Format" (SSF). A few days ago, the specification draft, a sample, and early source code were commited to his repository on sourceforge.

SSF is roughly based on Cascading Style Sheets (CSS), a technology who has common place usage on webdesign. It is very flexible, however, as you will see from the sample, it is extremely verbose and long, making it very impractical to edit manually. (In their defense, editing ASS/SSA styles manually is close to impossible, unless you memorize the argument order.)

A few days earlier, I have writen a very rough draft of a new format proposal, which I call AS5, which is a relatively simple-to-implement extension to ASS, adding several features which I consider vital. I have discussed it with a few friends (mostly on #aegisub), whom mostly liked the idea. equinox set up a wiki page based on the rough draft, so other people can contribute to the format (I don't intend to claim ownership of it, all I want is a better format available).

Most of you have probably also heard of "Universal Subtitle Format" (USF), an ASS replacement format developed by the CoreCodec team, and accepted by the Matroska Team as the container's official subtitle format. There is no official mention of this, but USF seems to be an XML version of ASS, with a few new features added. I haven't studied USF and SSF deeply, so I cannot tell which one is more advanced, but, from what I can see, SSF seems to have the higher hand. Unfortunately, since USF is XML, it's even more verbose and hard to manually edit than SSF is. USF specifications and samples can be downloaded here.

Now, seeing that a few fansubbers still use SSA (as opposed to ASS), - although many of them actually use ASS-only tags, such as \pos, into SSA - I wonder if any of those new formats would ever catch, assuming they had a working implementation.

And speaking of implementation, here are how things are going:

SSA/ASS: Full rasterizer implementation by vsfilter, and many full-featured editors are available: Aegisub, Sabbu, Medusa, SSA itself, etc. There is also a high number of tools available, such as SSATool.

AS5: Hacked up specs, no rasterizer implementation yet (ASA will probably support it, and it should be easy after it supports ASS), no support in any programs (though it will be very easy after it already supports ASS).

USF: Some horrible editor programs support it (i.e. ChronoSub, Medusa's worse-than-the-original successor), but no truly working rasterizer (one has been working on since the dawn of time, though).

SSF: No editor program supports it, but Gabest has already began work to make vsfilter support it, so that shouldn't take too long.


Now, the real issue with both USF and SSF is, while they're hard to read and write manually, they're even worse for a program to support. SSA/ASS are very simple and efficient formats, in particular, it's easy to edit subtitle content by text, so programs can just present that to user (indeed, that's what all of them do, although Aegisub and SSATool actually have to parse and comprehend every tag in order for a number of features to work).

To write a program that fully supports USF or SSF, while remaining user friendly is a daunting task, much harder than it is for ASS format. Naturally, this directly concerns me, since I am the Aegisub lead developer. From what I can tell of SSF, implementing good support for it in Aegisub would be extremely troublesome. I'm sure that kryptolus will think the same, regarding implementing it in Sabbu.

I'm posting because I would like to hear the opinion of the community in all those formats, and what do you expect of the future in this subject.


And thanks for bearing with my long rambling (if you've actually managed to do so )

[LytHka]

Well, the new generations will move on and try using the new formats, but us old hags will stick to ASS/SSA because it works and because it's simple. As you've said already, there is no real support for the new formats and when there will be new support, the developers are going to go each their own way with the applications. I assume this because you're the Aegisub developer and might implement AS5 support, I can't think of a reason why you would want to promote SSF, for example. Which ends us with a question: Do we really want a "war"?

I'm personally against this "war" because for fansubbers the SSA/ASS formats suffice. If you want to get graphical, you can use AFX. As a person who doesn't care about bit errors and format space usage and cares about what he *sees*, I think SSA/ASS will serve us good in the future as it has always. Everything else is an excess, IMO.

These new formats will need users if you want a "war," which leads me to think that we'll be seeing high class whoring of them, which will only disrupt the serenity of our beloved community and quite possibly further divide it.

I've just signed on the list of people who will ignore the new formats. Have fun moving the tombstones of digisubbing.

[ArchMageZeratuL]

Originally Posted by LytHka

I assume this because you're the Aegisub developer and might implement AS5 support, I can't think of a reason why you would want to promote SSF, for example.

I will implement any format that is useful and not a huge headache to implement. I wouldn't ignore other formats just because I helped develop another - that's the Microsoft way of things. Still, implementing AS5 is more likely than the others (assuming that it is ever supported by a rasterizer) because it's far simpler - in fact, it's not much more beyond an extension to ASS.

Originally Posted by LytHka

Which ends us with a question: Do we really want a "war"?

Of course not, but those formats are being developed anyway, so, in the end, we need to choose which one will each of us use.

Originally Posted by LytHka

I'm personally against this "war" because for fansubbers the SSA/ASS formats suffice. If you want to get graphical, you can use AFX. As a person who doesn't care about bit errors and format space usage and cares about what he *sees*, I think SSA/ASS will serve us good in the future as it has always. Everything else is an excess, IMO.

I think that, while SSA/ASS do many things very well, it misses some vital functionality, such as arbitrary transformations and gradients. AS5 solves both of those issues. The idea is that AS5 is to ASS what ASS is to SSA. Using After Effects is much more of a hassle, in comparison, so, if it could be entirely avoided, that would be the best in the opinion of many typesetters and encoders.

Originally Posted by LytHka

I've just signed on the list of people who will ignore the new formats. Have fun moving the tombstones of digisubbing.

I'm not trying to move anything, just asking for opinions. :P

[Quarkboy]

Although I'm all for innovation, it's sort of an unfortunate truth that a human readable/editable format is really important. Scripts need to be looked at at all steps in the process, and being tied to using a particular program to view the files forces EVERYONE in a group to learn and use that program.
I just had my editor complain to me about having to edit an ssa file... I can only fear what she'd do when I tell her she has to use some new program to edit the script .

But, I'm always open to newer better things. Can you be more specific about the advantages these formats have over ssa/ass?

Plus (and I'm going to sound like a mean old man here), having new "toys" in the format will encourage people to USE them... Making fancier karaoke and effects EASIER is not what fansubs need these days. I don't really like encouraging it.

[IRJustman]

LytH: I personally think there's always room for innovation. I'm by no means suggesting you hop on the bandwagon. Use what works for you. If you don't like what's being introduced, you're welcome to not use it. However, making snide comments serves no useful purpose.

--Ian.

[ArchMageZeratuL]

Originally Posted by Quarkboy

Although I'm all for innovation, it's sort of an unfortunate truth that a human readable/editable format is really important. Scripts need to be looked at at all steps in the process, and being tied to using a particular program to view the files forces EVERYONE in a group to learn and use that program.
I just had my editor complain to me about having to edit an ssa file... I can only fear what she'd do when I tell her she has to use some new program to edit the script .

Yes, I'm afraid that SSF/USF will suffer great resistance from that fact alone, if they ever get a proper rasterizer/editor.

Originally Posted by Quarkboy

But, I'm always open to newer better things. Can you be more specific about the advantages these formats have over ssa/ass?

As I've said, I didn't study the specs in detail, but they're linked in the post, if you want to do so... but make sure that you share the results with us.

SSF, at least, allows you to define your own custom tags and, apparently, do more advanced effects.

Originally Posted by Quarkboy

Plus (and I'm going to sound like a mean old man here), having new "toys" in the format will encourage people to USE them... Making fancier karaoke and effects EASIER is not what fansubs need these days. I don't really like encouraging it.

That IS a valid point (the Photoshop "bevel and emboss" effect is the ultimate proof of that), but I don't believe in removing freedom from users just because it will incite them to misuse it.

[Quarkboy]

Okay, here's the two features that I'd like to see in a new subtitle format that I'm sure isn't in these.

What I mean is, I want a tag which would make a sub anchored to the source video's background, which would move/resize based on the global motion/scale of the background video. This would make doing moving signs a cinch. Also add in the ability to overlay/dynamically scale a bitmap.
Add that and alpha overlaying (i.e. add in a zipped grayscale png for transparency overlay/ arbitrary clipping) and then there's no excuse to use afx anymore .

[LytHka]

Originally Posted by IRJustman

LytH: I personally think there's always room for innovation. I'm by no means suggesting you hop on the bandwagon. Use what works for you. If you don't like what's being introduced, you're welcome to not use it. However, making snide comments serves no useful purpose.

These weren't snide comments, it's the reality of this hobby. You don't need to learn another subtitling script language just to get a point across, or in our case, a translation. Developing another subtitling specs/format/parser just to make gradients is a complete waste of resources. People are just focusing too much on their tools and ways instead of actually subbing. I guess you're just bored, AMZ.

Originally Posted by lamer_de

Hence, i don't excatly see what this discussion is supposed to achieve.

As I've said, high class whoring, if a parser comes up

[lamer_de]

The point is that:

a) USF has been around for quite some time (a year, 2, don't know) and nobody is using it. There is a very high chance that it will stay that way.
b) SFF is unreadable to persons who don't program regulary. It supports some interesting features, but because everyone in the fansubbing community is using the ssa/ass format, it's highly unlikely a switch will occur.
c) AS5 has the advantage that it keeps the well known structure and adds some usefull functions, but because gabest has no interest in this (else he wouldn't do SSF), seeing an implementation of it is probably as likely as Duke Nukem Forever getting a release in the next month.
As already noted on the wiki, there were like 20 attempts to simply emulate the vsfilter capabilities and all failed, and AS5 is even more complex than that (3D rendering, free paths etc) [yes, asa has managed to display subs, but so has the ffmpeg project. The real stuff starts from here, I guess.].

Hence, i don't excatly see what this discussion is supposed to achieve. If there were working implementations then I could understand it, but so, it's just theoretical. There won't be a war, because there are no contestants :P

CU,
lamer_de

[Draders]

I really like how ASS is layed out, its simple and easy to edit in notepad as well as write programs for, also if one just extended ass to support new tags/features it would be easy to update previous programs to support the new tags (assuming they weren't coded badly :P), I think the bigger problem than ass lacking feaures is vsfilter being horribly slow to render things. Like you can easily do gradients in ASS but it just takes forever to render. It would be interesting for someone or a group of people to make their own version of vsfilter from the ground up taking advantage of hardware rendering and making it overal more efficent. Its pretty sad when it can take 10 seconds in vdub to render a frame of a karaoke. It seems to have more to do with the number of lines overlapping then the complexity of the lines.

[gumbaloom]

Why are people trying to make a subtitle parser do what the likes of Adobe After-Effects+Photoshop can do a whole lot quicker and easier.

You need to draw the line somewhere. I'm tended to agree with the people that say that SSA and ASS are perfectly fine and they can already accomplish the job very nicely. If you really want complex stuff just bite the bullet and go AFX.

If you try to make the subtitle renderer do all the complicated stuff all you're going to do is give the encoder a headache because of the scripts taking too long to process. I've seen it myself with even just simple things like gradients. How a simple 30k ssa can spiral in to an 8mb SSA and take 2 hours to render for just 20 seconds. I've also heard talk of other things like Live-eviL's Gilgamesh logo back in the day which took an inordinate amount of time to render properly for something that was so small.

Yeah I know theres loads of SSA purists who will stick up the cross and claim blasphemy at the thought of using AFX but we have to be realistic these days. Fansubbing is a hobby and we shouldn't be divesting loads of time like 2 hours in SSA doing a sign which could have been accomplished in a quarter of the time in AFX *JUST* because of a desire to remain a purist SSA typesetter. Move with the times people, realise that subtitle renderers are NOT in the same leage as photoshop and use AFX where it's appropriate.

Regards


-gumbaloom

[LytHka]

Originally Posted by gumbaloom

Move with the times people, realise that subtitle renderers are NOT in the same leage as photoshop and use AFX where it's appropriate.

Also, be sure to buy the program.

[Akirasuto.]

The ability to invoke a (optional) gui for coordinates would be nice.

I think an important task is what you, the would-be lead developer, wants to get from this potential project. Extending or rewriting the spec for the sake of doing so will ultimately fail, as evident from the numerous attempts on vsfilter. Best to stick the KISS principle in your goal, even when you want a more encompassing solution.

[jfs]

People (at least some) want softsubs. More and more fansubs do use softsubs, I believe. But Gabests renderer is dead slow and only works on Windows. Ergo, an alternative is needed.
I believe an alternative to TextSub is the number one thing needed right now. Something fast that still remains compatible. And indeed, several (too many) projects are already working on it.

But if implementing a new renderer hopefully with better speed and design (I can't tell if the TextSub design is good or bad, the code is simply too hard to understand properly) than TextSub, why not try to overcome some of the shortcomings of current SSA v4.00+ (or ++ if you want, it's not much of a change).
Personally I don't care much for actual calculations in subtitles, like SSF allows. I rather just want some simple, easy to read and easy to write, for both humans and machines, markup language. Nothing more, nothing less.
If people want advanced effects, they can sure figure out how to "misuse" it. People already are.

I don't believe people will pick up a new format just because it's new. It needs to be substiantially better. (As an example, this is why the switch to first XviD/DivX happened, and now to H.264.) Neither USF or SSF fulfill that. Both allow pretty much the same as SSA 4+, just with a more convoluted, hard to read and write by humans. Maybe easy to parse for rendering by machines, but it easily ends up as hard to present in a sensible format to a human who wants to edit it. And try making a machine writing optimal SSF code, I don't think you'll end up with anything more than SSA but with a more verbose syntax unless you want saving to take ages.

No, I believe the only natural path to take is really extending onto SSA, trimming it down as AS5 does. It's easy to read and write for both humans and machines, and while maybe lacking the structuredness of USF and the flexibility of SSF, it can still store something that renders the same images. And in the end, that's what counts, isn't it?

As for AFX and Photoshop, how useful are they for softsubbing? Yes, that's right, useless. And both are "payware" as I like to call it; I'm sure many people want to use free alternatives to commercial software whenever possible, especially when working on a non-profit hobby as fansubbing is. (And doesn't it have a nice ring to it when you can say "this professional-quality effect was made entirely using free software, in all three senses of free"?)

[ArchMageZeratuL]

Originally Posted by Quarkboy

What I mean is, I want a tag which would make a sub anchored to the source video's background, which would move/resize based on the global motion/scale of the background video. This would make doing moving signs a cinch. Also add in the ability to overlay/dynamically scale a bitmap.

Aegisub 1.10 supports that with Tentacle's motion tracker. You time the subtitles normally, then have the motion tracker run, select which points you want it to track, and it'll generate several instances of the line, with proper positioning and timing.

An unsupported 1.10 pre-release version can be downloaded at http://aegisub.cellosoft.com/prerel/. You only need the exe, or the Microsoft.VC80.CRT.rar if you need any errors, including the "wrong configuration" one. Also, an updated .chm is available, with a topic on how to use the motion tracker.

Originally Posted by lamer_de

c) AS5 has the advantage that it keeps the well known structure and adds some usefull functions, but because gabest has no interest in this (else he wouldn't do SSF), seeing an implementation of it is probably as likely as Duke Nukem Forever getting a release in the next month.

When Gabest shown the AS5 specs, he said that he had seen it already, and then, when asked what he thinks about it, he said "I don't". The specs were made known to him just soon after he released SSF, so he didn't know about AS5 when he started working on it. That means that there COULD be some chance that he's interested in supporting it, but, alas, now that he's working on SSF, those chances are small. Bad timing. :|

Originally Posted by Akirasuto.

The ability to invoke a (optional) gui for coordinates would be nice.

What do you mean?

[gumbaloom]

Originally Posted by jfs

As for AFX and Photoshop, how useful are they for softsubbing? Yes, that's right, useless. And both are "payware" as I like to call it; I'm sure many people want to use free alternatives to commercial software whenever possible, especially when working on a non-profit hobby as fansubbing is. (And doesn't it have a nice ring to it when you can say "this professional-quality effect was made entirely using free software, in all three senses of free"?)

As I predicted the Salem Witch Trials begin again!

Death to all ye heathens who have the gaul to suggest that using Adobe After Effects in fansubbing might actually be a good thing!.

*Gets out the boiling oil*

On a lighter note - if people want soft subs then why do they need a fancy subtitle specification that can render all sorts of flashy SFX?. If the reason given for a new spec is so we can have on the fly soft sub rendered exploding bread karaoke tell me where the nearest cliff is so I can jump off it please

-gumbaloom

[[darkfire]]

I would rather like an open source After Effects alternative. I don't know how that would be done seeing that afx is very complicated. But it would be nice if that was avaliable

[Bot1]

Originally Posted by gumbaloom

You need to draw the line somewhere. I'm tended to agree with the people that say that SSA and ASS are perfectly fine and they can already accomplish the job very nicely. If you really want complex stuff just bite the bullet and go AFX.

If you try to make the subtitle renderer do all the complicated stuff all you're going to do is give the encoder a headache because of the scripts taking too long to process. I've seen it myself with even just simple things like gradients. How a simple 30k ssa can spiral in to an 8mb SSA and take 2 hours to render for just 20 seconds. I've also heard talk of other things like Live-eviL's Gilgamesh logo back in the day which took an inordinate amount of time to render properly for something that was so small.

a new and better written rasterizer could at the same time make harsubbing stuff much easier also as a softsubber i want to be able to do as advanced an effect i can (not just for karaoke, also for logos and very difficulf signs) without having to use AFX or some weird workaround with what ass can do now that will be very advanced at have to be hardsubbed anyway

also this:

Originally Posted by gumbaloom

As I predicted the Salem Witch Trials begin again!

Death to all ye heathens who have the gaul to suggest that using Adobe After Effects in fansubbing might actually be a good thing!.

is a completely useless reply, he didn't say that AFX isn't usefull for fansubbing, he said it was useless for SOFTSUBBING and there is no way that you can't agree with that

[LytHka]

Originally Posted by Bot1

a new and better written rasterizer could at the same time make harsubbing stuff much easier also as a softsubber i want to be able to do as advanced an effect i can (not just for karaoke, also for logos and very difficulf signs) without having to use AFX or some weird workaround with what ass can do now that will be very advanced at have to be hardsubbed anyway

This is just another confirmation to gumbaloom's post about the softsub usage of this program -- it won't matter since the parser won't be able to render effects on the fly, or is AMZ perhaps as ambitious as to make that a reality? If you're smart, no matter what you use, ASS/SSA/AFX/future AS5/other shit future formats, you'll render video clips and force the encoder to put them in the video, if you're rendering SFX.

[Akirasuto.]

Originally Posted by ArchMageZeratuL

What do you mean?

Shows the x,y coordinate of user-selected pixel(s); helps with positioning signs and drawing clips. Such should be possible to via avisynth interface or virtualdub's, without the use of specialist apps like aegisubs, etc.