How should a new subtitle format look like?
 

What should a new format for subtitling have and have not?
Syntax & semantic - how structured?
What's good on ASS or others, what's bad or should just be change a little bit?

Please post your opinions and don't flame. This should be a constructive thread to help (maybe) developing a new format + renderer in future.

-----------------------------------------------------------------------

I begin:

• Short information section without rendering descriptions.
• Styles with text prefix.
• 3D capability (f.e. 3D position).
• Raw text rendering (without stripping spaces and tabulator accepted).

I think .ass pretty much does everything and Aegisub is all-singing, all-dancing. Why is there a need for a new format?

If you have to ask, you're not using ASS often or didn't understand the thread question.
F.e. a lot of typesetters are using AFX because ASS styling is too weak and working in plain text with ASS could surely be better.

I don't think there's a pressing need to add in advanced typesetting features to ASS, because once you talk about replacing AFX or even simulating some of its features you have to actually define a rendering specification.

The only feature I would like to have in ASS is support for more advanced timing formats:

1. Millisecond precision
2. Frame number based with framerate index to support VFR video subtitles. (i.e. subtitles have a start and end frame, and there's metadata that stores the framerates for each segment of the video).
3. Support for SMPTE timecodes and timecode resets. I.e. I'd like to be able to have a sub file with native support for SMPTE timecodes like 29.97 and 23.976 non-drop, plus a single sub file that can handle subtitles for videos where the timecode track resets in the middle.

You can already include pictures and movies in ASS. That means you can create any kind of subtitles you ever want.

It might use better positioning (i.e. I use 4:3 screen and move subs below the picture - and often signs look terrible; additional settings for different resolutions or better hinting might be a good solution) and timing settings as mentioned above, but certainly things like rendering are only limited by the subs creation software and the subbers skill and willingness to go to great lengths to achieve better effect.

When did ASS start opening pictures and movies? I can't find anything that does such a thing, unless you're talking about assdraw. Don't try to make a movie with assdraw...

I never said editing software had the necessary capabilities, but as for the format itself,
it has events: Dialogue, Picture, Sound, Movie

Although you'd probably have to do it manually

Yes per the document it supports those things, no software has ever supported those things in ASS. You could put in the lines, you would never get it rendered or played.
(Substation Alpha 4 did support those things, somewhat. But you would need a genlock to use that for subtitling.)

Doesn't change the fact that the specification has all those things, so format needs not upgrade in that regards.
I had the feeling the Picture event was supported, although it might've been those ASSDraw vector images instead.

The weirdest thing is that it even has Command support, which not only opens huge potential vulnerability (that's why no one in their right mind would implement that part), but also destroys crossplatform compatibility altogether.

The ASS (and SSA format on which it is based) has always been a specification defined by the software/hardware which implements it.
Anything other than [Dialog], [Comment] has never been implemented in any software to my knowledge, and most especially hasn't been implemented in vsfilter, which is the de-facto "reference implementation" (by being in fact, the only full-featured implementation of a renderer).

To say those commands are part of the standard means nothing other than that the text document of the standard lists those fields (and gives a rough description), but that text document is functionally nothing more than a ASS parsing syntax, not a specification. The true specification is only what is implemented by VSfilter. Anything else would count as an "extension to ASS" in my opinion.

Would be nice if there were a way to make the subs scroll, like the beginning of Star Wars movies. It's useful for audio only subs. The vid would generally be a still filler image.

...but you can do that pretty easily with .ass. {\move(640,-1000,640,0)\frx45} or something. The \frx might have to go in the other direction but otherwise, pretty basic.

I guess I should have explained myself better. What I'm looking for is this:
Lines would start at the bottom of the screen.
As each new line appears, it would push all the old lines up, until they disappear off the top.
The movement is not continuous: it only happens as new lines are said.

You can do this already by just modifying the default behavior for colliding lines:

In Aegisub go to File->Properties and change "Collision" from "Normal" to "Reverse".

Then just have all your lines overlap and it'll happen just how you want.

Gabest never got around to add reverse collisions to VSFilter. As far as I'm aware, nobody else ever did either.

The way VSFilter resolves collisions is simply keep track of what lines were visible on the previous frame renderes, and those lines' display positions, then for all lines still visible on the current frame, keep their positions, and try to place all new lines based on that.

You can get the effect of reverse collisions by hardsubbing the file in reverse to an intermediate lossless (the encoder must pull out the frames in reverse order!), then reverse that lossless again afterwards to the final encode.
Of course that gives you a hardsub.
(Otherwise you'd have to have some script to hard-resolve the collisions, making every subtitle placed with \pos or layers+margins.)

Some other ideas:

• animation acceleration as math expression (f.e. \t(500,-1000,sin(t*PI),\pos(100,200)) )
• stenciling to define areas by drawing something for later rendering in- or outside
• transformation stacking for 3D positioning (with depth test)
• texturing with texture coordinates, edge mode and color multiplication/masking

A keyboard button to add leadin and leadout at the same time, and another to do scene-timing for one specific line on the "fly" by simply pushing another keyboard button, likewise, another keyboard button for "linking" lines (lines with gaps between them)

The discussion is about a new format, not about a new editor.

I think what would be most useful would be support for non-linear movement :D
A nice gimmick would be support for vertical gradients in texts.

It's a bit old thread, but I guess its OK to post here.

IMO it would be good to have multiple subtitle streams in a single file *&* an ability to show them at the same time. Why is this useful? Well, fansubers place a lot of different stuff in subs, like: comments from TLs, TLed signs, karaoke etc. For some, only karaoke will be enough, for others only main text etc. Of course it can be done by placing multiple subs into multimedia container, but it has it's own disadvantages. In any case, if not in the format spec, it would be good at least to see this as a feature of the renderer.

I have some other thoughts about new format, but not sure whether they will be worth implementing, they are mostly a nice neat stuff.

+1 to SinsI: I also think that Advanced SSA is a good format that need relatively slight change, e.g. support for picture/video embedding, some workaround for margins (margins set in style can be overridden only if they are non-zero).

Also, it seems that people have forgotten that Medusa/Adv. SSA author hadn't completed this format. E.g. you can find in Medusa install dir prototype of karaoke templates similar to styles, which could be also implemented:
Though, as a hobbyist programmer I can say that making all-around format is a bad thing to do. For example look at SVG (Scalable *Vector* *Graphics*), which can contain raster graphics, fonts (well, they're almost vector graphics, so...) and video (in v2, I guess). Once it was a language to describe vector graphics, nowadays it had become closer to HTML... Thus, it's better if we either stick with basic format like SubRip etc., or go straight for some multimedia description language.