How should a new subtitle format look like?

[Quarkboy]

Personally I believe that some effort should be put into TTML (Timed Text Markup Language)

http://www.w3.org/TR/2013/WD-ttaf1-dfxp-20130131/

which at least has a draft standard from WC3. It already includes most of the features of .ass and that have been requested here and would have the theoretical benefit of being supported by web browsers if it ever gets fully approved.
I'm not sure there's any software that implements more than a tiny subset of it at this point, though.

But wouldn't it be great to have browser html5 natively support subtitle display with all the fancy options?

[8day]

I've just re-read this topic and... have anyone noticed that it's only about typeset? Almost no one here had said nothing about improvement that concerns subtitles, everything was about the so called "view".

In that case, you guys really should work on some post-processing framework rather than subs. Even HTML-like subs probably won't be enough. VapourSynth seems to be getting more mature with every day, so it may be a good choice. Or, use Blender which will do all that you want, esp. with VapourSynth.

P.S. Sorry for somewhat trollish post, but isn't lots of stuff that you want is already possible in Blender? Why not create format that describes only subs (collision handling, timing etc), use it as a Model and feed it to Blender (which is already cross-platform & mature) which will do all View-related stuff?

[rarely_upset]

Maybe I'm just uninformed about VapourSynth and Blender, but I think we're talking softsubs and you're talking hardsubs.

[Youkakun]

Originally Posted by 8day

I've just re-read this topic and... have anyone noticed that it's only about typeset? Almost no one here had said nothing about improvement that concerns subtitles, everything was about the so called "view".

In that case, you guys really should work on some post-processing framework rather than subs. Even HTML-like subs probably won't be enough. VapourSynth seems to be getting more mature with every day, so it may be a good choice. Or, use Blender which will do all that you want, esp. with VapourSynth.

Right, this discussion isn't (only) about typesetting, it's about subtitle format design.
Anyway doing all this 'view' stuff additionally in video editing software requires hardsubbing and more work. I like to write some text and put my tiny script with others together for the final overlay.

Originally Posted by Quarkboy

Personally I believe that some effort should be put into TTML (Timed Text Markup Language)

http://www.w3.org/TR/2013/WD-ttaf1-dfxp-20130131/

...
I'm not sure there's any software that implements more than a tiny subset of it at this point, though.

Many people have good ideas for new formats, but the software implementation is always the real problem. I've already started to write a new renderer, but stopped by the effort to find a fitting to-memory-rendering library or write a complete renderer by myself.

[Quarkboy]

To step back for a bit into theory, what is really needed is a standard with multiple defined layers...

So for example at the simplest layer you just have absolute times and text, with no formatting or display requirements, and then the next layer adds some things like alignment, line breaks, italics, and other simple things, and then the next layer adds fonts, borders, coloring, etc etc...

Then you can have subtitle renderers which you can define as properly implementing the standard "up to level X"....

That way you have a format which is always at least playable, and your software will conform to the standard up to all the features covered by the level it is rated at.