ISSUE-447 (Other types of track are possible): Mapping is only for presentation, i.e. subtitle and captions, and other types of WebVTT document are possible [TTML WebVTT mapping note]

ISSUE-447 (Other types of track are possible): Mapping is only for presentation, i.e. subtitle and captions, and other types of WebVTT document are possible [TTML WebVTT mapping note]

http://www.w3.org/AudioVideo/TT/tracker/issues/447

Raised by: Nigel Megitt
On product: TTML WebVTT mapping note

Whereas a single TTML file may be structured to include metadata and chapter information, through the use of <div> elements and the ttm:role attribute and ttm:metadata for example, the common usage pattern for WebVTT e.g. in HTML5 text tracks, is to reference a different separate WebVTT file for each of those usages, some of which are intended for UA display, and others intended to be hidden.

See http://www.w3.org/TR/html5/embedded-content-0.html#timed-text-tracks for HTML 5 text track definition.

The mapping document as it stands looks like it is only intended for mapping TTML to WebVTT where the intention is to display the contents as subtitles or captions, so it should state that scope clearly. Content differences should be added to §1.3. §2 should possibly be renamed "TTML to WebVTT for captions or subtitles". §3 should possibly be renamed "WebVTT captions and subtitles to TTML". 

There certainly needs to be a section that outlines the semantic differences in how different kinds of content can be or are expected to be contained in each format.

A further option is to define a different mapping for the equivalent of text track kind = "chapters" or "metadata" or even possibly "descriptions".

If we do want to support those alternate types of data here are some more thoughts:

See http://www.w3.org/TR/ttml1/#metadata-attribute-role and https://www.w3.org/wiki/TTML/RoleRegistry for TTML role information.

In the case of chapters I would expect there to be one chapter per div in the TTML file, and for any content labelled as ttm:role="title" to be carried into the WebVTT file. The overall timing of the div should be calculated, and any content with a different role discarded.

In the case of descriptions I would expect to map only content marked as ttm:role="description" to be carried into the WebVTT file.

In the case of metadata it's unclear what best to do - maybe specify that a user defined scheme should be used to translate contents of ttm:metadata into some WebVTT construct?

Thinking about the reverse case of WebVTT to TTML mapping, a similar thing needs to be done. The mapping strategy needs to take into account what sort of content is expected in the WebVTT file, which I suspect can not be derived from within the WebVTT itself (help me out if I'm wrong here!).

Mapping WebVTT chapters to TTML should result in a set of timed <div> elements marked as ttm:role="title" with appropriate content. This is significantly different to subtitles or captions, where the expectation is to create <p> elements. Sophisticated mapping implementations should not be prohibited from taking in 2 WebVTT files, one for chapters and the other for displayed content, and creating a single TTML file whose div elements correspond to the chapters and whose p elements correspond to the displayed subtitles or captions and are located appropriately for their timing within whatever div they fall within.

Received on Friday, 25 September 2015 16:38:30 UTC