Audio description failure technique #4390
Conversation
Addresses #1768 Creates a failure technique to show the conditions under which a video fails 1.2.5 due to important visual information not being described in pauses in the dialogue
✅ Deploy Preview for wcag2 ready!
To edit notification comments on pull requests, go to your Netlify project configuration. |
I've removed "must" to address Scott's concern
Remove note styling and add in wording suggested in #3806
I think it may look better with the note. Trying to retain just for the pre-existing paragraph to see how that looks
techniques/failures/F113.html
Outdated
| <p>For each occurrence of synchronized time-based media:</p> | ||
| <ol> | ||
| <li> Check that all important visual information has been conveyed in the narration.</li> | ||
| <li> Check that narration has been added in all appropriate pauses in the dialogue. </li> |
There was a problem hiding this comment.
Check that audio description is provided for any visual information necessary to understand the content which is not conveyed by audio in the synchronized media.
There was a problem hiding this comment.
This second check is there to make sure any available appropriate pauses are used.
The first check has already validated that the audio descriptions either cover all the material, or they don't.
Here, we're testing that all the pauses have been used for narration.
Together they give us the gates for the failure:
#1 passes= no failure
#1 fails and #2 passes=no failure
#1 fails, #2 fails=failure
There was a problem hiding this comment.
The original working is actually checking that all possibly useful pauses are used, which may not be appropriate.
| <p>This describes a failure condition for all techniques involving audio descriptions. Audio descriptions can either be provided as part of the soundtrack's original narration or be added to audio during pauses in existing dialogue. If important actions, characters, scene changes, and on-screen text are only conveyed visually, appropriate pauses in dialogue need to be used to provide this information as audio descriptions.</p> | ||
| <p class="note">Not all pauses are usable for audio descriptions. If the pauses in dialogue are too short (for instance, less than 2 seconds), or do not occur in proximity to the visual content that needs to be described, they may not be appropriate for audio descriptions.</p> | ||
| <p>In situations where important visual information is being conveyed at the same time as non-spoken audio (such as music and sound effects), the technique of "audio ducking" can be used. This involves dropping the overall sound level so that it is easier to distinguish the narration that is added during pauses in dialogue.</p> | ||
| <p>This technique can work well with background music and sounds, but audio ducking has the potential to mask important audio information. Often such audio information can convey much of the sense of what is visually happening. There may be some pauses in dialogue where non-spoken audio information is so important that the addition of audio descriptions is unnecessary or inappropriate.</p> |
There was a problem hiding this comment.
Having trouble following the point of this paragraph. Is it that ducking can obscure needed audio information or that sometimes the non-spoken audio can eliminate the need for some descriptions?
There was a problem hiding this comment.
I assume you mean the last one?
Is it that ducking can obscure needed audio information or that sometimes the non-spoken audio can eliminate the need for some descriptions?
A bit of both, I think. There is material about not filling every pause with more information if the audio is actually more important than the video
For instance, see example 2
Some_ examples would probably be useful, although maybe that would be better in a sufficient technique than in a failure technique.
There was a problem hiding this comment.
@awkawk, let me know if this helps clarify. Open to suggestions!
| <p>This technique can work well with background music and sounds, but audio ducking has the potential to mask important audio information. Often such audio information can convey much of the sense of what is visually happening. There may be some pauses in dialogue where non-spoken audio information is so important that the addition of audio descriptions is unnecessary or inappropriate.</p> | |
| <p>This technique can work well with background music and sounds, but audio ducking has the potential to mask important audio information. Sometimes such audio information can convey much of the sense of the video. </p> | |
| <p class="note">There may be some pauses in dialogue where non-spoken audio information is so important that the addition of audio descriptions is inappropriate.</p> |
Incorporating the suggestion
|
Here's what I was thinking this was testing: For all synchronized media with video:
Is this what you are saying? |
Nope.
If you fail both, you fail. I was hoping that was obvious from the title "Failure of Success Criterion 1.2.5 due to not using available pauses in dialogue to provide audio descriptions of important visual content" |
Ha! Hope is dashed! :) The title is not the test procedure, you can't rely on it. So you want to combine my 1 and 2 into 1. How about:
both true: pass |
|
That wording of the second check seems harder to parse, to me. I've update the wording to make it more what you wanted. Can you explain where you are primarily snagging on this now?
If both of these are false, the failure applies. |
You really seem to want to have audio description in all pauses, but this is definitely the snag. |
I would not say that. What the test is doing is checking if all the important video is described. If it's all described, then unused pauses are immaterial, it passes.
|
|
@mbgower I emailed with Bryan Gould, who is at WGBH-NCAM and worked for many years as a description editor at WGBH-DVS. Bryan replied to me (quoted sections are from my email to Bryan, non-quoted are his replies):
I agree. For 2, the word “appropriate” provides enough ambiguity that this check may very well be interpreted as a “fill-every-pause” mandate. If that is the case, the next question is, what constitutes a pause? A lot of information can be conveyed with a single well-placed word, for example, “Later” or “Elsewhere.” These words can be vital to understanding but very few one-second breaks in dialogue require description. The practice of filling every pause with description may also be detrimental. In order to reduce cognitive load, the viewer may filter out the “excess” description. As a result, the viewer may miss the actually important information.
Correct. A description editor’s main role is to apply context to all visual information in order to deem what is “important” and thus requiring of description. For example, a 10-second view of an orchid requires a different description approach depending on whether the context is botany, photography, or part of a montage set to music. There are also instances when a pause in dialogue is intended for the viewer to reflect on what was just said, this is especially true in education and training videos. In other words, there is almost always too much visual information to describe and the description editor is always deciding what is important. Finally, dedication to context helps to prevent the cognitive load issue raised above. |
|
Thanks for providing that context and outside, informed opinion in your latest comment, @awkawk. There's nothing stated there that is new to me, and I share all those concerns and observations based on my own limited experience in this field. However, this is pretty similar, philosophically, to problems that exist in 1.1.1 for images. In the case of 1.1.1, we realistically have a gross litmus test: "is there an alt (and if not, is this purely decorative)?" This is the basic check every automated checker currently does. And then we have a qualitative test: "is it equivalent?" We've all seen some pretty awful alt text. But concerns with the quality of the alt has not deterred us from insisting that alt text must exist.
That is a fundamental challenge with the art of audio descriptions. The pauses, ultimately, do to a great degree determine what one decides are the important visuals. (I even wrote a blog about this very topic once.) I intentionally incorporated the idea of "appropriate" into the check for pauses to help provide some flexibility there. For audio descriptions, we only have the text we were given to work. I thought a gross test for 1.2.5 could be "Is there any audio description?" That seems entirely supported by the normative text. But several people have pushed back against that idea. The only other gross tests we seem to have at our disposal with the normative wording are:
Both of these are highly subjective. I sympathize with the concerns Bryan states, but the reality is that as a percentage of total videos on the web, the number that are audio described is much closer to 0% than 1%. I think we have to provide some kind of gross test for audio descriptions before we turn to worrying about whether the descriptions are good or not. Rejecting usable pauses as a metric for 1.2.5 for fear someone creates crappy audio descriptions seems unfortunate to me. I've already opened an issue to create a sufficient technique that documents how to do good audio descriptions. To me, that is a better way of addressing his concern than ruling out "pauses" as a part of a failure assessment. |
|
@awkawk a compromise is possibly to remove "all" from both of the checks . I'm not keen on it (since it really lowers the bar for passing) BUT it would create a higher bar for failing, which I believe is what you're advocating for?
Follow up: Or maybe the "all" only needs to go on the second check? |
|
After some discussion with @awkawk , we tentatively arrived at: Test Procedure
If checks 1 and 2 are false, then this failure condition applies and the content fails the success criterion. Note that I have only updated the original PR language here. I haven't incorporated any of the other suggestions to date. |
|
Moved to ready for approval; will move back depending on @awkawk response. |
There was a problem hiding this comment.
I don't think the 2nd step of test procedure tracks closely enough to the failure name and it's not as simple as it could be. I recommend:
For each occurrence of synchronized media containing video:
- Check that all important visual information has been conveyed in the narration.
- Check that there are available pauses in dialogue where narration could be added.
If check 1 is false and check 2 is true, then this failure condition applies and the content fails the success criterion.
Incorporated based on discussion with @awkawk
The PR was updated after your comment, and may have at least partially addressed. Please review and comment. |
Closes #3806
Addresses #1768
Creates a failure technique to show the conditions under which a video fails 1.2.5 due to important visual information not being described in pauses in the dialogue
Preview