Caption synchronization tolerance

An open issue:

>67. Tolerance of synchronizing captions and audio descriptions
>
>Checkpoint 1.2 in the 14 August 2001 draft has the following success 
>criterion and note:
>
>3. descriptions and captions are synchronized with the events they 
>represent to within a tolerance of X. [Note: We need to research the 
>tolerance. Any information on this is appreciated.]

Meeting minutes say:

>#67WC ask Geoff Freed or someone at WBGH.
>JW GV didn't want a number.
>Action WC: Ask Geoff.

While we're waiting for Geoff, I will point out that a specific time 
unit cannot be given. The best you can say is "descriptions and 
captions are closely synchronized" or "reasonably synchronized."

In offline or prerecorded captioning, it is not uncommon to see 
captions when nobody is speaking-- e.g., a wordy but 
quickly-delivered monologue from Xaviera just ends, and we the 
audience are now presented with the stunned expression of Miguel, who 
cannot believe the diatribe he was just subjected to. We use this 
moment of silence to catch the captions up to what were actually 
spoken. I have witnessed two two-line captions appearing in such a 
space. That puts the last word of the last caption about six seconds 
after verbal delivery. No problem.

One also sees captions in advance of delivery, though not very often, 
and I won't bother giving examples.

Typically, though, in pop-up captioning the words are already on the 
screen as they are being spoken.

Now, the pressing issue is real-time captioning, where captions can 
unavoidably appear a full half minute from the original utterance. 
That is very uncommon, but it happens:

* captioning software crashes

* program returns from commercial unexpectedly fast and 
stenocaptioner is busy blowing her nose ("OMIGOD! It's back!")

* captioner must manually spell an important word, then wait for the 
software to spit it out (and spelling must be done by entering, in 
one technique, the syllables corresponding to the name of the letter: 
ell owe you gee aitch bee owe are owe you gee aitch [wait] [proceed])

* item was previously captioned; captioner recognizes it as such; 
captioner goes to scroll up the prepared captions; captions cannot be 
found; "OMIGOD"; keep looking; either find the file or begin 
stenocaptioning again

* captioner is using modified and very weird sports-captioning 
technique, in which words are captioned in real time, the words are 
gathered up into pop-on captions, then sent out.

Even under the best conditions, you can wait three seconds for a word 
to appear after it is spoken. In French, you can wait ten seconds, 
for reasons I won't bore you with.


In audio description, you also find descriptions heard in advance of 
or after the fact. There is a tendency to avoid enormous time 
differentials, preferring to step on the dialogue and other audio to 
describe something in the here and now. This happens more often than 
neophytes think, though of course the incompetent Canadians will do 
anything to avoid that phenomenon, going so far as to refuse to 
describe something rather than describe over dialogue. In either 
case, there is a bias against extreme delays or anticipation in 
description timing.

The most important fact? *All* of these variations are acceptable or 
unavoidable. You can find several of them in a single day's viewing.

Do not infer from these anecdotes that 30 seconds would be an 
adequate time threshold for captioning. There are too many variables 
at work to assign a specific number. Just say "closely" or 
"reasonably" and trust people not to screw it up. Of my many 
complaints with captioning and my occasional complaints with 
description, unacceptably-delayed or -early presence isn't even on 
the Top 50. There are too many good reasons for slow or fast captions 
or descriptions for a body like the WCAG, whose members spend very 
little time watching accessible TV, to set deceptively tidy numerical 
standards.
-- 
         Joe Clark | joeclark@joeclark.org | <http://joeclark.org/access/>
         Accessibility articles, resources, and critiques ||
           "I can't pretend to understand the mind of Joe Clark"
           -- Larry Goldberg

Received on Thursday, 23 August 2001 20:07:49 UTC