02 Jul 2013

WebVTT Voice Tags

For a while we've had a bug open to update WebVTT voice tags to show the text of the speaker in the way that the spec wants us to. At the time I opened the bug I thought that this was something that we would need to be doing as that was my current understanding. As the bug has progressed though I've discovered a number of things.

The first was that WebVTT voice tags aren't actually supposed to have a default display. I originally thought that they would as it makes sense to me that they would. Why have voice tags if they don't show anything? However, after a number of discussions on the public text track mailing list, where Simon Pieters was kind enough to help me, I found out that the original reason for this was so that authors could turn voice tags on and off if they want and control how they display. It would be harder to do this if the author had to struggle with some kind of CSS that we've already applied to the tag.

For example if the author wanted to apply custom CSS to the voice tag it may be as easy as this:

::cue(v):before { 
    content: attr(voice) ":  "; 

Which would show: Title Here: Subtitle Here

The second thing that I realized stemmed from the issue above. That is that GetCueAsHTML() is not actually what the spec would have us use to render the subtitles on the video. This is the way we're currently doing it in Firefox. Rather, it would like us to generate CSS boxes, not elements, and use those to render the subtitles on the video instead. The reasoning here is that CSS boxes are more light weight and you might want that when updating the video every 250 ms, heh. However, Ralph has since told me that we can just use the same code in GetCueAsHTML() and the processing model as that's an internal functioning of the spec which we don't need to follow exactly. Possibly in the future we can switch over to using just straight up CSS boxes, if Gecko allows you to do such a thing, in order to get some performance boost or something. We shall have to see.

GetCueAsHTML() is primarily for showing the cue in other parts of the webpage or for custom subtitles as it's apart of the exposed API. Whereas, the processing model generates a bunch of CSS boxes with positioning and styling applied to them specifically for rendering on the video as subtitles.

It's been hard to keep up with the changing spec. When we first sat down to write the browser side GetCueAsHTML() was the only area for us to hook into; the processing model wasn't there, I'm fairly certain. So for a while it's been stuck in my mind that this is how it should work. I'm glad that this all came to light as I can move forward with a better understanding of the spec.

Until next time.

- Tagged in mozilla, seneca college, open-source, and cdot.

comments powered by Disqus

Wow coding Copyleft Richard Eyre 2013

Code for this site can be found on GitHub.