RE: WebVTT

>My question was how to have center aligned paint-on captions without moving the painted on text as it appears. But I think Glenn has answered that by saying you have both display and visibility to choose from. If you choose to set visibility, however, you have to have the whole text in the render tree - are you certain that the TTML model with successive cues allows for that?

Yes. If you use visibility:hidden, then the invisible text is accounted for in the layout, so it will not appear to reflow, whereas display:none never shows up in the tree at all, but the render model copes with both. In both cases it is an entirely different render tree in each cue.

>I don't quite follow, but if TTML doesn't define a rendered and relies on CSS for the HTML rendering case, I'd be surprised if there wasn't a simple way.

Not sure what you are saying here exactly, but if you have a different proposal I'd be happy to hear it, the current model however works, so there's no real reason to change.

>Right, if that's what you're worried about, you can always convert a VTT cue with time stamps into multiple TTML cues separated at the time stamp. I thought we were talking about converting from TTML to VTT.

I am considering both directions.

-----Original Message-----
From: Silvia Pfeiffer [mailto:silviapfeiffer1@gmail.com] 
Sent: 17 June 2013 10:55
To: Sean Hayes
Cc: John Birch; public-tt@w3.org
Subject: Re: WebVTT

On Mon, Jun 17, 2013 at 7:42 PM, Sean Hayes <Sean.Hayes@microsoft.com> wrote:
>
>>So you represent Karaoke differently from paint-on captions?
> Well it depends on what you want it to look like, it would be possible to represent it the same way if you chose to style the display property manually for paint on.
>
>>Also, how do you deal with the rendering problem when the future text is not in the render tree. For example, you render the first word in a centered cue. If the other words are not yet present in the render tree, the first word is rendered in the middle. Then successively the words are added, which means the text has to get re-balanced and the first word move successively to the left.
>
> There is no rendering problem here if you model as we do. Each cue is simply a static piece of HTML. You take one out and replace it with the next. Your observation is however why paint on styling is left justified, But yes it is possible to do text as you describe in TTML, and a myriad of other possibilities..

My question was how to have center aligned paint-on captions without moving the painted on text as it appears. But I think Glenn has answered that by saying you have both display and visibility to choose from. If you choose to set visibility, however, you have to have the whole text in the render tree - are you certain that the TTML model with successive cues allows for that?


> Which is precisely why it's not possible to map general TTML to the wavefront model. To generalize it I think would be to have a mini TTML renderer in each cue, which then begs the question of the cue model at all, so I don't think we'll go there.

I don't quite follow, but if TTML doesn't define a rendered and relies on CSS for the HTML rendering case, I'd be surprised if there wasn't a simple way.


>>Is the extra functionality useful for HTML? If not, it sounds like the "wavefront" model would be sufficient.
> That's not really the right question unfortunately. The TTML author has the ability to write it, so the question is how to model it. It seems we aren't worried about the events, so I guess we'll just accept that converting to VTT from TTML will generate less events.

Right, if that's what you're worried about, you can always convert a VTT cue with time stamps into multiple TTML cues separated at the time stamp. I thought we were talking about converting from TTML to VTT.

Cheers,
Silvia.

Received on Monday, 17 June 2013 10:07:05 UTC