Are you one of those people that get all mad when we translate ojou-sama and the like? Because this isn’t the group that supplicates itself to weeaboos.
Is there any reason why this anime’s intro and devil survivor 2 intro’s are lagging on my computer? Pretty sure it has to do with the typesetting and encoding, but I’ve never had trouble before with commie’s encodes.
1. Update your playback (CCCP is what you should be using and the only thing we support)
2. Demux the subs and remove the typesetting
3. Get a new computer
Because?
1) Non-Standard ass stuff?
-> don’t use non-standard stuff.
2) crappy/nonexistent implementation of new ass stuff?
-> inb4 breaking some libass’ devs balls.
3) idk. Pls. herkz explain with some detail what you/commie did to let libass shit bricks.
libass renders the tags differently than xy-vs. So when I stack on thousands of lines of softsub typesetting, things are bound to be breaking sooner or later.
Because libass has a lot of bugs and it is not being maintained any longer. Not to mention xy-vsfilter is considerably faster now which is relevant too.
Bummer (the bugs). I don’t know there isn’t much activity but, last update was Oct. 2012. I wouldn’t dismiss this as abandoned. …Ok now I understand why you chose xy-filter. According to http://code.google.com/p/libass/wiki/IssuesAndDifferences, vsfilter also has a few issues
eg: > BiDi support in VSFilter is broken beyond repair. libass does not emulate this.
The difference is the way libass renders some ASS tags in combinations that we often use is completely wrong and makes the typesetting unreadable. That’s pretty much the only thing that matters for fansubbing and it is a massive disadvantage.
Sorry, can’t reply below you: > The difference is the way libass renders some ASS tags in combinations that we often use is completely wrong and makes the typesetting unreadable. That’s pretty much the only thing that matters for fansubbing and it is a massive disadvantage.
Sounds like ccs2/3 IE8-ish problem. So it is in fact case 2). :(.
Well, as long as the problem is only the intro. Would be hilarious if I cannot watch my animu after university/work without windows.
Yes, but that has been and is still broken, making libass basically unusable for fansubbers. It doesn’t matter until it and any other bugs like it are fixed. So you can say “Oh well, it’s being worked on,” but that doesn’t really change the fact that it’s still broken.
> The latest patch was both created and pushed just ten days ago. The latest update on the issue tracker was posted just a week ago.
I’m talking about official releases here, not “unstable”/”dev”/”whatever” commits in the gitrepo.
Fortunately for me, I use arch linux, so I just need to install libass-git from the AUR, but as herkz already mentioned, the bugs he’s talking about aren’t fixed anyway, yet.
What I don’t really understand is why are the libass devs teaking vsfilter as a baseline. Isn’t there an “official” ASS specification?
I’m kinda impressed, or rather irritated that ASS is used everywhere, although there isn’t really a specification. Afair, ASS is an extension to SSA (ok .. as it seems its SSA v4+ or sth like that). I guess, SSA also doesn’t have an “official” spec.
I agree libass is not perfect. I agree xy-VSFilter should be used for watching anime when possible.
But libass is not useless either. You make it sound like even the most basic sign typesetting is incompatible with libass, but of course that’s not true.
By the way, what do you use that shear resetting thing (which issue #80 is about) for? I do believe VSFilter is the de facto standard and should be emulated however insane it is, so I’m definitely not saying libass should stick with its current behaviour here, but I do think its current behaviour makes sense and VSFilter’s doesn’t. I’m just struggling to understand why you would actually want any text property change to reset your baseline after shearing.
{\an7\pos(244,490)\fscy100\fs30\fay-0.48\fscx55\fnFranklin Gothic Demi\blur0.75\c&H292626&\frz305.3\t(0,7840,\blur1)}It's so funny, I was surprised! \NI'm so anxious to know what \Nhappens, I can't return to the \Nafterlife!
I hope you can see how easy it is to break libass. Just using \fax or \fay once and then putting \N or \h in the line anywhere will result in incorrect behavior in libass (unless you actually think it should work how it does).
edit: It doesn’t render the borders correctly either
I’m not sure about the border though. The only show I’m up to date with is Precure, so which episode of which series is this? I can download it and test, and potentially fix the border bug in libass if there is one.
Oh, I think I see it by the pixels in the screenshots. VSFilter seems to add the borders after doing the rotation, whereas libass seems to do it the other way round and rotate the whole thing including the borders. libass’ behaviour makes more sense to me; do you disagree? Of course, both renderers should behave the same, but the question is which one to fix.
Ok cool, I love this action atm. My suggestion is to document all the unclear behaviour that differs from vs-filter (eg: because you think the vs-filter behaviour is not right etc) and some commments why it should differ.
The other way (as already mentioned) would be to emulate vs-filter as far as possible. Tbfh, I don’t know what would be the better approach.
> In the VSFilter version, the border is suspended in mid-air.
Doesn’t sound like intended, or supposed to be to me.
I feared that something like that would haven. Libass must “emulate” vsfilter bugs or it gets really ugly sooner or later (discriminating Linux users like me :().
> Doesn’t sound like intended, or supposed to be to me.
Indeed, and that’s my point. Thanks for agreeing with that. But herkz says the other version, which puts the border on the book cover, looks wrong to him.
I understand why people want VSFilter and libass to produce identical output. I don’t understand why people want output that makes no sense. Particularly here: isn’t the point of typesetting a sign to make it look like it’s part of the picture drawn by the animation studio? Surely you then want your borders to lie in the same plane as the text and the surface the text is written on.
@Cortos-2 (if you still follow this thread) >> Doesn’t sound like intended, or supposed to be to me.
>Indeed, and that’s my point. Thanks for agreeing with that. But herkz says the other version, which puts the border on the book cover, looks wrong to him.
I don’t really know the structure of libass, but I have a suggestion regarding the vs-filter bugged feature xy, vs. supposed to be feature xy.
How about a idk, compatibility/quirks layer.
The “supposed-to-be” version is already implemented.
How about a quirks mode, which emulates this buggy border behaviour.
As long as vsfilter hasn’t fixed this bug, the quirks mode should be active.
Eg:
BorderBlahImplementationFactory.createImplementation(IS_QUIRKS_MODE_ACTIVE).
Or whatever the best solution you come up with.
I can’t say I haven’t considered this myself, but it doesn’t seem like a good idea however I look at it.
Do you mean that for offering selectable behaviour or just for keeping the current code around in case VSFilter’s behaviour changes in the future? If the latter, reverting git commits is easy enough, so it would be cleaner to just make libass use VSFilter borders and drop the current code (even though I think this would mean dropping quite a lot of code, which might end up producing conflicts when reverting later).
If the former, a significant problem with such a quirks mode is that unlike HTML which has doctypes, ASS has no way of deciding which scripts should be rendered in quirks mode and which in non-quirks. There’s not much point letting the user choose either: it’s not like the user knows any better (and most users won’t even know what this means at all). Not to mention adding a quirks mode for just this border issue, which nobody seems to even have noticed before now, seems excessive.
I’m going to try to survey some typesetters on IRC.
> Do you mean that for offering selectable behaviour or just for keeping the current code around in case VSFilter’s behaviour changes in the future? If the latter, reverting git commits is easy enough, so it would be cleaner to just make libass use VSFilter borders and drop the current code (even though I think this would mean dropping quite a lot of code, which might end up producing conflicts when reverting later).
I meant the former. And I pulled this example out my a**.
I don’t know how many bugs there are, but if they are negligible (this translation of überschaubar doesn’t sound right to me), I would say a set of variables which are set to true or not, and depending on the value, libass uses the quirks mode for a specific feature or not. > There’s not much point letting the user choose either: it’s not like the user knows any better (and most users won’t even know what this means at all).
It doesn’t need to be the user.
It could be done via a comment in the ASS script that gets evaluated by libass
(is there some comment syntax in ASS?).
Something similar is done in Javascript (ECMA Strict mode).
// Whole-script strict mode syntax
“use strict”;
var v = “Hi! I’m a strict mode script!”;
Based on this:
//use_border_quirks_mode=true
This could be even incorporated in typesetting software, so that the typesetter chooses sane values (as long as the typesetter isn’t a complete n00b, he should understand what that means, hopefully), or just default the values to a sane setting.
That’s the trouble: even this conversation started not because VSFilter is impossible to fix but rather because typesetters stubbornly stick to its bugs.
That said, I think I understand your suggestion better now, and I must say I’ve had a similar idea before—no, in fact, I still harbour it. There’s been talk of a new subtitle format for years now… er, more like, there was talk of a new subtitle format back in 2008, and I’ve heard nothing new about it since then. Since this new format based on ASS is not happening, an alternative I see is just ASS but with a formal specification that makes sense. So it would still need a new version number not to conflict with existing scripts, but the only real changes would be fixing VSFilter bugs and nonsensical pseudofeatures. Although frankly, I don’t remember the new format, according to the draft specification I read, being that much different from ASS, so I’m not sure if it makes sense not to go all the way and finalize that specification.
For reference, after a somewhat positive response on IRC, I’ve filed a bug with xy-VSFilter. Don’t expect it to be fixed any time soon, but at least it can be said that libass is doing it right this time.
Also, this is late, but I feel I should clarify my previous comment a bit: I was afraid that nobody would use an extension introduced by libass. However nice the extension may be, maintaining code that isn’t used by anyone is ultimately a waste of effort.
@bug report, cool thanks. > Also, this is late, but I feel I should clarify my previous comment a bit: I was afraid that nobody would use an extension introduced by libass. However nice the extension may be, maintaining code that isn’t used by anyone is ultimately a waste of effort.
You mean regarding my proposed “meta-language” in the comment section?
You are probably right. I also don’t know where this should be communicated to so that it might be adapted in the future.
Although, this is just monkey patching anyway. The only right step would be to abandon ASS and create a new specification based on ASS – completely documented with a reference implementation and for each feature which seems problematic, some comment why it does do things in “this” specific way… Just sayin’.
Oh, here’s a thing. If you want to copy the quirks mode suggestion to a place where it’s more easily found, I guess you can use the libass issue tracker on Google Code. But you’ll probably get a negative response from Grigori or wm4 and may see the issue closed as wontfix. It still will be more easily found though.
Maybe I will try that, but I guess the reason why you didn’t do this in the first place is, because those 2 guys you mentioned just scrapped the idea because it’s a bad idea in their opinion or something like that.
Because libass and xy-VSFilter render some stuff differently. Some discrepancies are due to VSFilter bugs that nobody dares to fix any more (where libass does the thing that makes sense because it was written without knowledge of the particular VSFilter bug). Some are due to fundamental differences in the way libass and VSFilter do things; most if not all of these can be called libass bugs for some definition of bug. Some are due to VSFilter delegating stuff to Windows components whose working principles are not known in enough detail. Of course, as any piece of software, libass is also not safe from the most ordinary kind of bugs. Oh, and there’s also bidirectional text handling, which is apparently either horribly broken or non-existent in VSFilter, so the leading libass developer believes he’s really doing it right and it really should be fixed in VSFilter.
Specific libass problems are known to its developers and are getting fixed; this just happens very slowly.
In the case of SnK I can understand the lag from the typesetting of the title (Episode 1 was 90k lines Episode 2+ is about 16k). However if you are lagging on the ds2 intro (1.9k lines)I suggest upgrading to the latest cccp. If you are already on the latest cccp then I suggest that you either
A) Feed your gerbil powering that thing you call a computer.
B) get a new pc.
Make sure that gerbil is fed. I also have an i7 and can run the ds2 op no problem. The SnK opening I slightly stutter for a second and then it runs fine.
I can only assume it’s your playback. Check to make sure you have xy-vsfilter working properly.
No one might be ever reading this, but I found the bottleneck. See, I stream from my NAS and one part of the intro manages to cap my 10 mbps wireless link. This is slightly ridiculous.
incidentally, make sure you uninstall all other codec packs. I’ve recently had this problem, and updating to CCCP at first didn’t seem to change it… except my old k-lite was still installed. deleting both and reinstalling cccp removed those issues (of lag)
I’ve been having issues with subs for this and Yahari. For this, the subs display properly during the OP, but once the episode starts, no subs. Other recent releases are fine.
Correction: Although I believe I haven’t changed anything in CCCP except updating, even previous Commie releases have sub issues. Episodes I was able to play last week properly now have no subs, even though they’re enabled. Very weird…
FINALLY a name we all can agree on is the superior.
Indeed
wasn’t it Eoten on Onslaught
Damnit go work on Mahoyo!
Great work :D
completely unrelated things but okay
that size ! decreased
however, good work commie :D
i still wait for Railgun S 2
dem eoten are bck
as my name says.
keep out! >[
Thanks for the subs guys! :D
Thanks comi, great works.
Not sure if you care, but:
http://i.imgur.com/qFpNM0z.jpg
deeply
Also,
16:24 I believe it should be breach;
16:34 “The livestock were were”.
the editor is being whipped as we speak
with cream?
;)
Substituting “breech” for “breach” is indefensible and I have no excuse. My apologies.
I still respect your work. As they say, it’s the trust between people.
more troll names?
they’re titans, not trolls
I think you mean eotens
Doesn’t already set the official name Attack on Titan?
oh yeah, eoten? idk , confused…”Titan” will much more familiar with this stuff…
the splash title animation cause too much lag…
you’ve got some reading to do
oh yeah forgot to mention it…
on OP theme before closing, it’s Gurren no Yumiya, “Crimson Arrow” not Gurren no Jager “Crimson Hunter”
<Rotpain> that’s deliberate
well if it’s intentionally i can’t complain because this is free fansubs, sorry for expecting too much…
What is that supposed to mean?
i mean for seriously paying attention to your works all this time…
Are you one of those people that get all mad when we translate ojou-sama and the like? Because this isn’t the group that supplicates itself to weeaboos.
i agree. i don’t mind that but better don’t mislead information within translation. IT’s OK as long it’s close with the original…
but you’re doing it with wrong meanings…but please not with self satisfaction, i hope next episode will be better, i looking forward for it too.
No seriously what the fuck are you people talking about? I can’t understand what you and the other guy want through your completely broken English.
( ¬‿¬)
Hummmm. That intro…
“Sie sind das Essen und wir sind die Jäger”
“They are the food and we are the Hunters”
…might be an important point since some people are calling everyone sheltered livestock.
no shit sherlock
protip: main character’s last name is also Jäger
Lol. Just remembered this:
http://themavesite.com/TMS-Pictures/2011-12/SieSehenMichRodeln.jpg
Is there any reason why this anime’s intro and devil survivor 2 intro’s are lagging on my computer? Pretty sure it has to do with the typesetting and encoding, but I’ve never had trouble before with commie’s encodes.
In b4 “Get a better computer”…
Typesetting overload. hdr has declared war on xy-vsfilter
I see, thanks for the info. May I suggest adding a way to remove the typesetting? Like another subtitle track or something like that?
No idea if it’s possible as I’ve never done typesetting before, but it sure would be nice when typesetting just starts eating too much cpu ressources.
1. Update your playback (CCCP is what you should be using and the only thing we support)
2. Demux the subs and remove the typesetting
3. Get a new computer
That’s about all you can do.
>CCCP is what you should be using and the only thing we support
Not everyone is using Windows, you know.
well you’re shit out of luck if you aren’t because libass definitely will not work on some of the typesetting
Because?
1) Non-Standard ass stuff?
-> don’t use non-standard stuff.
2) crappy/nonexistent implementation of new ass stuff?
-> inb4 breaking some libass’ devs balls.
3) idk. Pls. herkz explain with some detail what you/commie did to let libass shit bricks.
>3) idk. Pls. herkz explain with some detail what you/commie did to let libass shit bricks.
We let Haidaraaaaa loose into the world.
implying either 1)
or 4) Commie’s at fault. Which semantically is the same as 1).
Either way, now I know at least whom to blame.
libass renders the tags differently than xy-vs. So when I stack on thousands of lines of softsub typesetting, things are bound to be breaking sooner or later.
Because libass has a lot of bugs and it is not being maintained any longer. Not to mention xy-vsfilter is considerably faster now which is relevant too.
>not being maintained any longer
lolwut
there’s a bunch of bugs that have existed for at least a year that aren’t fixed so…
Some known bugs existing and not getting fixed does not mean it’s unmaintained when other bugs are getting fixed.
Bummer (the bugs). I don’t know there isn’t much activity but, last update was Oct. 2012. I wouldn’t dismiss this as abandoned. …Ok now I understand why you chose xy-filter. According to http://code.google.com/p/libass/wiki/IssuesAndDifferences, vsfilter also has a few issues
eg:
> BiDi support in VSFilter is broken beyond repair. libass does not emulate this.
Guess, libass should be forked too.
The difference is the way libass renders some ASS tags in combinations that we often use is completely wrong and makes the typesetting unreadable. That’s pretty much the only thing that matters for fansubbing and it is a massive disadvantage.
Sorry, can’t reply below you:
> The difference is the way libass renders some ASS tags in combinations that we often use is completely wrong and makes the typesetting unreadable. That’s pretty much the only thing that matters for fansubbing and it is a massive disadvantage.
Sounds like ccs2/3 IE8-ish problem. So it is in fact case 2). :(.
Well, as long as the problem is only the intro. Would be hilarious if I cannot watch my animu after university/work without windows.
> The difference is the way libass renders some ASS tags in combinations that we often use is completely wrong and makes the typesetting unreadable.
First, make sure you’re using the latest libass. Next, either there’s an open issue on http://code.google.com/p/libass/issues/list or it didn’t happen.
https://code.google.com/p/libass/issues/detail?id=80 This is the main one that I know of that actually comes up pretty often. Well this, and all the related issues.
herkz, note how that bug is relatively actively being investigated.
Basement Dad, I’ve just noticed this:
> but, last update was Oct. 2012.
The latest patch was both created and pushed just ten days ago. The latest update on the issue tracker was posted just a week ago.
Yes, but that has been and is still broken, making libass basically unusable for fansubbers. It doesn’t matter until it and any other bugs like it are fixed. So you can say “Oh well, it’s being worked on,” but that doesn’t really change the fact that it’s still broken.
> The latest patch was both created and pushed just ten days ago. The latest update on the issue tracker was posted just a week ago.
I’m talking about official releases here, not “unstable”/”dev”/”whatever” commits in the gitrepo.
Fortunately for me, I use arch linux, so I just need to install libass-git from the AUR, but as herkz already mentioned, the bugs he’s talking about aren’t fixed anyway, yet.
What I don’t really understand is why are the libass devs teaking vsfilter as a baseline. Isn’t there an “official” ASS specification?
>Isn’t there an “official” ASS specification?
If only.
I’m kinda impressed, or rather irritated that ASS is used everywhere, although there isn’t really a specification. Afair, ASS is an extension to SSA (ok .. as it seems its SSA v4+ or sth like that). I guess, SSA also doesn’t have an “official” spec.
Well no one has made anything better yet sadly.
Seems like that. :(. Too bad I don’t know anything about that rendering stuff.
<- Console Jockey here.
I agree libass is not perfect. I agree xy-VSFilter should be used for watching anime when possible.
But libass is not useless either. You make it sound like even the most basic sign typesetting is incompatible with libass, but of course that’s not true.
By the way, what do you use that shear resetting thing (which issue #80 is about) for? I do believe VSFilter is the de facto standard and should be emulated however insane it is, so I’m definitely not saying libass should stick with its current behaviour here, but I do think its current behaviour makes sense and VSFilter’s doesn’t. I’m just struggling to understand why you would actually want any text property change to reset your baseline after shearing.
http://i.imgur.com/gaUWpJe.jpg
http://i.imgur.com/2r002ut.jpg
First is vsfilter, second is libass.
Here’s the line in the script:
{\an7\pos(244,490)\fscy100\fs30\fay-0.48\fscx55\fnFranklin Gothic Demi\blur0.75\c&H292626&\frz305.3\t(0,7840,\blur1)}It's so funny, I was surprised! \NI'm so anxious to know what \Nhappens, I can't return to the \Nafterlife!I hope you can see how easy it is to break libass. Just using \fax or \fay once and then putting \N or \h in the line anywhere will result in incorrect behavior in libass (unless you actually think it should work how it does).
edit: It doesn’t render the borders correctly either
Oh boy, I should’ve reloaded the page before hitting Submit.
> I’m talking about official releases here, not “unstable”/”dev”/”whatever” commits in the gitrepo.
But talking about official tagged releases is pretty useless.
> as herkz already mentioned, the bugs he’s talking about aren’t fixed anyway
That’s far from the only relevant bug, don’t you think? Currently the git HEAD contains quite a lot of important bug fixes on top of the last release.
herkz, this is why I asked you to make sure you were using the latest libass. I have your line rendered just fine here.
I’m using whatever is included in Aegisub, obviously.
Even the latest VLC positions the line correctly. This bug was fixed in libass 0.10.1: http://code.google.com/p/libass/issues/detail?id=46&can=1, http://code.google.com/p/libass/source/detail?r=7af78014635dbe81c3328405aa68dd2cfef94bc4. And yeah, I agree this bug was pretty important. I was going to list is as an example of why you should use the git HEAD for Basement Dad until I saw it was already included in the last release.
I’m not sure about the border though. The only show I’m up to date with is Precure, so which episode of which series is this? I can download it and test, and potentially fix the border bug in libass if there is one.
It’s from Mangirl! episode 12.
Awesome.
Indeed, the border is rendered just as in your screenshot. I have some bug hunting to do!
Oh, I think I see it
by the pixelsin the screenshots. VSFilter seems to add the borders after doing the rotation, whereas libass seems to do it the other way round and rotate the whole thing including the borders. libass’ behaviour makes more sense to me; do you disagree? Of course, both renderers should behave the same, but the question is which one to fix.the libass version looks wrong to me, yes
Ok cool, I love this action atm. My suggestion is to document all the unclear behaviour that differs from vs-filter (eg: because you think the vs-filter behaviour is not right etc) and some commments why it should differ.
The other way (as already mentioned) would be to emulate vs-filter as far as possible. Tbfh, I don’t know what would be the better approach.
In the libass version, the border is on the book cover. In the VSFilter version, the border is suspended in mid-air.
But oh well.
> In the VSFilter version, the border is suspended in mid-air.
Doesn’t sound like intended, or supposed to be to me.
I feared that something like that would haven. Libass must “emulate” vsfilter bugs or it gets really ugly sooner or later (discriminating Linux users like me :().
> Doesn’t sound like intended, or supposed to be to me.
Indeed, and that’s my point. Thanks for agreeing with that. But herkz says the other version, which puts the border on the book cover, looks wrong to him.
I understand why people want VSFilter and libass to produce identical output. I don’t understand why people want output that makes no sense. Particularly here: isn’t the point of typesetting a sign to make it look like it’s part of the picture drawn by the animation studio? Surely you then want your borders to lie in the same plane as the text and the surface the text is written on.
herkz, if you’re still checking this thread for new comments, can you compare these two lines in VSFilter and libass?
{\pos(640,200)\fs128\frx70\bord10\c&HFFFFFF&\3c&H000000&}lololol\Nlololol\Nlololol\Nlololol\Nlololol\Nlololol\Nlololol\Nlololol\Nlololol\Nlololol\Nlololol\Nlololol\Nlololol\Nlololol\Nlololol\Nlololol\Nlololol\Nlololol\Nlololol\Nlololol\Nlololol\Nlololol\Nlololol\Nlololol\Nlololol\Nlololol\Nlololol\Nlololol\Nlololol\Nlololol\Nlololol\Nlololol\Nlololol\Nlololol\Nlololol\Nlololol\Nlololol\Nlololol\Nlololol\Nlololol\Nlololol
{\pos(640,300)\fs128\frx89\bord10}lololol\N
Add some \frx animations or something if you like. Do you still think VSFilter’s version is better? If so, why?
I wanted to add screenshots, but I somehow broke my Windows player setup in the process. Sorry.
@Cortos-2 (if you still follow this thread)
>> Doesn’t sound like intended, or supposed to be to me.
>Indeed, and that’s my point. Thanks for agreeing with that. But herkz says the other version, which puts the border on the book cover, looks wrong to him.
I don’t really know the structure of libass, but I have a suggestion regarding the vs-filter bugged feature xy, vs. supposed to be feature xy.
How about a idk, compatibility/quirks layer.
The “supposed-to-be” version is already implemented.
How about a quirks mode, which emulates this buggy border behaviour.
As long as vsfilter hasn’t fixed this bug, the quirks mode should be active.
Eg:
BorderBlahImplementationFactory.createImplementation(IS_QUIRKS_MODE_ACTIVE).
Or whatever the best solution you come up with.
I can’t say I haven’t considered this myself, but it doesn’t seem like a good idea however I look at it.
Do you mean that for offering selectable behaviour or just for keeping the current code around in case VSFilter’s behaviour changes in the future? If the latter, reverting git commits is easy enough, so it would be cleaner to just make libass use VSFilter borders and drop the current code (even though I think this would mean dropping quite a lot of code, which might end up producing conflicts when reverting later).
If the former, a significant problem with such a quirks mode is that unlike HTML which has doctypes, ASS has no way of deciding which scripts should be rendered in quirks mode and which in non-quirks. There’s not much point letting the user choose either: it’s not like the user knows any better (and most users won’t even know what this means at all). Not to mention adding a quirks mode for just this border issue, which nobody seems to even have noticed before now, seems excessive.
I’m going to try to survey some typesetters on IRC.
> Do you mean that for offering selectable behaviour or just for keeping the current code around in case VSFilter’s behaviour changes in the future? If the latter, reverting git commits is easy enough, so it would be cleaner to just make libass use VSFilter borders and drop the current code (even though I think this would mean dropping quite a lot of code, which might end up producing conflicts when reverting later).
I meant the former. And I pulled this example out my a**.
I don’t know how many bugs there are, but if they are negligible (this translation of überschaubar doesn’t sound right to me), I would say a set of variables which are set to true or not, and depending on the value, libass uses the quirks mode for a specific feature or not.
> There’s not much point letting the user choose either: it’s not like the user knows any better (and most users won’t even know what this means at all).
It doesn’t need to be the user.
It could be done via a comment in the ASS script that gets evaluated by libass
(is there some comment syntax in ASS?).
Something similar is done in Javascript (ECMA Strict mode).
// Whole-script strict mode syntax
“use strict”;
var v = “Hi! I’m a strict mode script!”;
Based on this:
//use_border_quirks_mode=true
This could be even incorporated in typesetting software, so that the typesetter chooses sane values (as long as the typesetter isn’t a complete n00b, he should understand what that means, hopefully), or just default the values to a sane setting.
Ok, gotcha.
[Script Info]
; use_border_quirks_mode=true
; use_blah_quirks_mode=false
; …whatever
Something like that.
That’s the trouble: even this conversation started not because VSFilter is impossible to fix but rather because typesetters stubbornly stick to its bugs.
That said, I think I understand your suggestion better now, and I must say I’ve had a similar idea before—no, in fact, I still harbour it. There’s been talk of a new subtitle format for years now… er, more like, there was talk of a new subtitle format back in 2008, and I’ve heard nothing new about it since then. Since this new format based on ASS is not happening, an alternative I see is just ASS but with a formal specification that makes sense. So it would still need a new version number not to conflict with existing scripts, but the only real changes would be fixing VSFilter bugs and nonsensical pseudofeatures. Although frankly, I don’t remember the new format, according to the draft specification I read, being that much different from ASS, so I’m not sure if it makes sense not to go all the way and finalize that specification.
For reference, after a somewhat positive response on IRC, I’ve filed a bug with xy-VSFilter. Don’t expect it to be fixed any time soon, but at least it can be said that libass is doing it right this time.
Also, this is late, but I feel I should clarify my previous comment a bit: I was afraid that nobody would use an extension introduced by libass. However nice the extension may be, maintaining code that isn’t used by anyone is ultimately a waste of effort.
@bug report, cool thanks.
> Also, this is late, but I feel I should clarify my previous comment a bit: I was afraid that nobody would use an extension introduced by libass. However nice the extension may be, maintaining code that isn’t used by anyone is ultimately a waste of effort.
You mean regarding my proposed “meta-language” in the comment section?
You are probably right. I also don’t know where this should be communicated to so that it might be adapted in the future.
Although, this is just monkey patching anyway. The only right step would be to abandon ASS and create a new specification based on ASS – completely documented with a reference implementation and for each feature which seems problematic, some comment why it does do things in “this” specific way… Just sayin’.
Indeed.
…I don’t think I can add anything useful to that.
Oh, here’s a thing. If you want to copy the quirks mode suggestion to a place where it’s more easily found, I guess you can use the libass issue tracker on Google Code. But you’ll probably get a negative response from Grigori or wm4 and may see the issue closed as wontfix. It still will be more easily found though.
Maybe I will try that, but I guess the reason why you didn’t do this in the first place is, because those 2 guys you mentioned just scrapped the idea because it’s a bad idea in their opinion or something like that.
Because libass and xy-VSFilter render some stuff differently. Some discrepancies are due to VSFilter bugs that nobody dares to fix any more (where libass does the thing that makes sense because it was written without knowledge of the particular VSFilter bug). Some are due to fundamental differences in the way libass and VSFilter do things; most if not all of these can be called libass bugs for some definition of bug. Some are due to VSFilter delegating stuff to Windows components whose working principles are not known in enough detail. Of course, as any piece of software, libass is also not safe from the most ordinary kind of bugs. Oh, and there’s also bidirectional text handling, which is apparently either horribly broken or non-existent in VSFilter, so the leading libass developer believes he’s really doing it right and it really should be fixed in VSFilter.
Specific libass problems are known to its developers and are getting fixed; this just happens very slowly.
extract with mkvtoolnixgui2, open with aegis, remove unnecessary line,
lag only occurs on OP song theme though….
In the case of SnK I can understand the lag from the typesetting of the title (Episode 1 was 90k lines Episode 2+ is about 16k). However if you are lagging on the ds2 intro (1.9k lines)I suggest upgrading to the latest cccp. If you are already on the latest cccp then I suggest that you either
A) Feed your gerbil powering that thing you call a computer.
B) get a new pc.
There has to be something else then, that gerbil is a 1st gen i7 :/. Anyhow, I’ll test my cpu usage with the vid running and find the bottleneck.
Make sure that gerbil is fed. I also have an i7 and can run the ds2 op no problem. The SnK opening I slightly stutter for a second and then it runs fine.
I can only assume it’s your playback. Check to make sure you have xy-vsfilter working properly.
No one might be ever reading this, but I found the bottleneck. See, I stream from my NAS and one part of the intro manages to cap my 10 mbps wireless link. This is slightly ridiculous.
It’s probably the title. It’s about 90K lines of clipped gradient.
The bottleneck is not processor. For me at least it’s the vram. But yeah, 90k was quite the overkill.
incidentally, make sure you uninstall all other codec packs. I’ve recently had this problem, and updating to CCCP at first didn’t seem to change it… except my old k-lite was still installed. deleting both and reinstalling cccp removed those issues (of lag)
Guys, just a question a bit off…
The subs of The World God Only Knows, both season 1 and 2, are Crunchyroll’s?
Thanks.
improved, but yes.
I’ve been having issues with subs for this and Yahari. For this, the subs display properly during the OP, but once the episode starts, no subs. Other recent releases are fine.
I’m using the latest CCCP (4/20/13).
Check the CRC of the file. It may not have downloaded correctly.
Correction: Although I believe I haven’t changed anything in CCCP except updating, even previous Commie releases have sub issues. Episodes I was able to play last week properly now have no subs, even though they’re enabled. Very weird…
Haali splitter issue? I switched to LAV splitter and subtitles appear now. I’m stumped…
No idea. Works for me. It’s difficult to diagnose a problem on someone else’s computer.
LOL! ive been w8ing for that till now, didnt noticed uve changed the name :P
I feel you missed a good opportunity with this, you could have linked to They Might be Giants – Boss of Me Now during the heat of the drama.
You might enjoy:
https://www.youtube.com/watch?feature=player_embedded&v=PDHCLjuOzXY
Sorry, this is even better.
http://www.youtube.com/watch?v=AC2okqCGMg4
unanimeted, could you please add to the examples in your TS manual how did you mananage to do this absolutely brilliant title TS?
He didn’t do the title TS.
With 20MB of typesetting
SOMEBODY PLEASE TELL LEGOFGUY TO MAKE A VERSION OF THIS.
-> Came here to download anime.
-> Got caught up reading the comments.
-> Had to come back just to download the torrent.