Re: Systematic access to media/plugin metadata

On Apr 12, 2011, at 1:13 PM, Eric Carlson wrote:

> 
> On Apr 12, 2011, at 12:20 PM, Laura Carlson wrote:
> 
>> 
>>>> How does the end user obtain the transcripts if there is no script?
>>>> 
>>> They can't, but as far as I know there is no standard way to mark a text
>>> track in an audio or video file as a "transcript"
>> 
>> Maybe there should be?
>> 
>  First, meta-data is stored and interpreted differently in different media file formats so this would have to be done for every format. Therefore I it is out of scope for the HTML WG.
> 
>  Secondly, I am not sure this would be especially useful because 
> 
  Oops, sent this too soon. I meant to say:

   Secondly, I am not sure this would be especially useful because an in-band transcript could not be downloaded separately from the rest of the video file.

eric




>>> so a script with
>>> hard-coded knowledge about the contents of such a movie will be required for
>>> this anyway.
>>> 
>>> In any case all of the samples in a media file are typically intermixed,
>>> eg. a text track will be broken up into small chunks and spread throughout
>>> the file. This means that it isn't usually possible to load only a text
>>> track,
>> 
>> So are you saying that most captions are not  just a text file with
>> time stamps? Is that correct?
>> 
>  Captions are text samples with timing information, but they do not necessarily come out of text files. Some media container formats can contain text samples (we have been calling these "in band" captions in our WG discussions) as well as audio and video data. For example if you have ever looked at a movie or video podcast with closed captions on an iPhone or iPad, those captions come from the video file, not a separate text file, and are being rendered QuickTime.
> 
>  I assumed that you were talking about "in-band" captions because your original email talked about accessing EXIF, IPTC, and XMP metadata from inside of image files.
> 
> 
>>> and in your example of a user with a slow network connection it will
>>> be necessary to download the entire video file even if they only want the
>>> transcript.
>> 
>> Downloading an entire video file would not be an option for a user on
>> a slow connection. Just starting a video grinds everything to a halt.
>> 
>  Right, that is why I mentioned that I didn't think in-band transcripts would be useful for your use case.
> 
>> When I asked Silvia for a transcript of her WebVTT explained Video
>> [1]. She kindly linked to
>> a full transcript of the described video in the form of a WebVTT file
>> [2]. That worked. Is there anyway that transcripts like this could be
>> extracted and offered to users on slow connections automatically?
>> 
>  I would imagine that whatever mechanism we agree on for linking to external long descriptions would work for transcripts as well - as long as the descriptions are not in the video file.
> 
> eric
> 
> 
>