Audio description failure technique

[mbgower]

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

[netlify]

✅ Deploy Preview for wcag2 ready!

Name	Link
🔨 Latest commit	e8f5216
🔍 Latest deploy log	https://app.netlify.com/projects/wcag2/deploys/6835ced83501000008f3486f
😎 Deploy Preview	https://deploy-preview-4390--wcag2.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.

---

To edit notification comments on pull requests, go to your Netlify project configuration.

[scottaohara]

just one suggestion. otherwise this makes sense to me - and the other delving into the applicability or essential nature of a pause can be handled in content to be added to the understanding doc

[awkawk]

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.

[mbgower]

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

[awkawk]

The original working is actually checking that all possibly useful pauses are used, which may not be appropriate.

[awkawk]

Here's what I was thinking this was testing:

For all synchronized media with video:

1. Check that all important visual information in the video is conveyed by the main soundtrack.
2. Check that any important visual information in the video that isn't conveyed in the main soundtrack is conveyed by audio description.

Is this what you are saying?

[mbgower]

Here's what I was thinking this was testing:

For all synchronized media with video:

1. Check that all important visual information in the video is conveyed by the main soundtrack.
2. Check that any important visual information in the video that isn't conveyed in the main soundtrack is conveyed by audio description.

Is this what you are saying?

Nope.

1. Check that all important visual information in the video is conveyed by the main soundtrack.
2. Check that all pauses are used for audio descriptions

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"

[awkawk]

Nope.

1. Check that all important visual information in the video is conveyed by the main soundtrack.
2. Check that all pauses are used for audio descriptions

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:

1. Check that all important visual information in the video is conveyed by the main soundtrack or by an audio description track.
2. Check that the main soundtrack has pauses that would be appropriate to provide audio description of important visual content.

both true: pass
both false: pass
1 true, 2 false: pass
1 false, 2 true: fail

[mbgower]

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?

1. Check that all important visual information is conveyed by the main soundtrack or by audio descriptions.
2. Check that audio descriptions have been added in all appropriate pauses in the dialogue.

If both of these are false, the failure applies.

[awkawk]

Can you explain where you are primarily snagging on this now?

You really seem to want to have audio description in all pauses, but this is definitely the snag.

[mbgower]

Can you explain where you are primarily snagging on this now?

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.
If all important visuals are not described, then check that all appropriate pauses have been used. If the meaningful visuals aren't all described and if all the appropriate pauses aren't utilitized, then it fails, because it hasn't met the SC (emphasis on all):

Audio description is provided for all prerecorded video content in synchronized media.

[awkawk]

@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’ve responded to your comments below:

My concern is the second check, which suggests that the right way to handle audio description is to fill every available pause with audio description if there is any information deemed important that is missing. I worry that we will get into a situation where well-intentioned people will think that there is some important information from the video that isn’t voiced and therefore must fill every pause.

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.

I suspect that even the best-described video might fail this evaluation, as there is way too much visual information to completely reproduce in AD, and there are times where a description editor needs to make a judgement about the impact on the mood of the content and viewer’s experience.

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.

[mbgower]

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.

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.

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:

• are there important visuals with no audio equivalent?
• are there pauses which could be used to insert audio descriptions?

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.

[mbgower]

@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?

1. Check that all important visual information has been conveyed in the narration.
2. Check that narration has been added in all appropriate pauses in the dialogue.

Follow up: Or maybe the "all" only needs to go on the second check?

[mbgower]

After some discussion with @awkawk , we tentatively arrived at:

Test Procedure
For each occurrence of synchronized media containing video:

1. Check that all important visual information has been conveyed in the narration.
2. Check that narration has been added in pauses in the dialogue, where appropriate.

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.

[mbgower]

Moved to ready for approval; will move back depending on @awkawk response.

[bruce-usab]

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:

1. Check that all important visual information has been conveyed in the narration.
2. 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.

[mbgower]

@bruce-usab

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.

The PR was updated after your comment, and may have at least partially addressed. Please review and comment.

[bruce-usab]

Check that audio descriptions have been added in pauses in the dialogue, where appropriate.

Is not as clear as:

Check that there are available pauses in dialogue where narration could be added.

1. The "where appropriate" is more ambiguous than "are available pauses".
2. Still retains the words "audio description" which has overloaded incorrect implications that there is a second audio track or distinct voice actor for the addition narration.

[GreggVan]

@bruce-usab Can you clarify your point? I'm not sure I understand.

To help - my confusion is around

1. The "where appropriate" is more ambiguous than "are available pauses".

the "where appropriate" relates to whether or not Audio Description is needed at all at that point in the video. The "are available pauses" relates to the opportunity to add it even if there was a need.

2. Still retains the words "audio description" which has overloaded implications that there is a second audio track or distinct voice actor for the addition narration.

Audio description is the normal term - and talks about adding description in gaps. Like captions - it can be the only audio track or an additional audio track (it is usually the latter). Do you think there is another term that should be used?

PS on a different topic (Sort of) I like your "test steps" suggestion further above

[bruce-usab]

PS on a different topic (Sort of) I like your "test steps" suggestion further above

IMHO, that's the important bit!

the "where appropriate" relates to whether or not Audio Description is needed at all at that point in the video

Understood. For an author/auditor this requires subjective judgement.

The "are available pauses" relates to the opportunity to add it even if there was a need.

Understood. Noting if there "are available pauses" is an objective observation.

Audio description is the normal term

Understood, and the glossary notes are entirely consistent with normal usage of the term.

Do you think there is another term that should be used?

Nope, not at all!

[mbgower]

@bruce-usab this is essentially your wording with me swapping out "avaiable" with "appropriate". I also (previously) swapped out "narration" for "audio descriptions".

Suggested change

	<li> Check that audio descriptions have been added in pauses in the dialogue, where appropriate. </li>
	</ol>
	</section>
	<section class="results"><h3>Expected Results</h3>
	<ul>
	<li>If checks 1 and 2 are false, then this failure condition applies and the content fails the success criterion.</li>
	<li>Check that there are appropriate pauses in dialogue where audio descriptions could be added.</li>
	</ol>
	</section>
	<section class="results"><h3>Expected Results</h3>
	<ul>
	<li>If check 1 is false and check 2 is true, then this failure condition applies and the content fails the success criterion.</li>

[mbgower]

I'll note that we've abandoned the use of "narration" in the checks. This occurred because once we included the normative wording in check 1 about "not understood from the main soundtrack along", we have covered any narration that exists in the main soundtrack. So we are only focused on additive audio descriptions in both checks.
I think for a failure technique this works fine.

[mbgower]

Suggested change

	<p>For each occurrence of synchronized media containing video:</p>
	<p>For each occurrence of synchronized media containing audio and video:</p>