Fediverse @lemmy.world Jupiter Rowland @hub.netzgemeinde.eu 1y ago

Alt-text: How detailed does it have to be?

This is going out to both the #Threadiverse and, because I can't keep this from happening, the rest of the #Fediverse where I've mentioned this issue before three months earlier.

In brief: I'm still not sure how much #AltText is optimal. And I tend to run into situations in which alt-text that describes everything in a picture will grow longer than any of you could possibly imagine in their wildest dreams.

Here's my situation:

I don't have a problem with writing a lot. Unlike most of you, I'm not on a phone. I'm on a desktop computer, and if I'm not, I'm on a laptop. I've always got a full-blown hardware keyboard, and I can touch-type with ten fingers. And I like to rant.
I'm on #Hubzilla. This means virtually no limit in post length and especially virtually no limit in alt-text length. The only limiting factor would be how much alt-text the instances where my posts are viewed can display. #Mastodon has a hard cap at 1,500 characters, for example.
I'm not the one to skimp on #accessibility rules unless they're technologically impossible for me to follow. I'd rather do too much than too little. This includes full transcriptions of all texts in a picture unless privacy issues speak against it, or unless I've got no way to source the original of a text anymore, and said text in the picture is ineligible even for me. Yes, I transcribe text that's one pixel high if I can get the original.
When I post pictures, I don't always post them Instagram/Pixelfed-style, i.e. posts that are about this particular picture. Instead, I often use pictures to illustrate the post. Hubzilla gives me all necessary means to write full-blown blog posts with all bells and whistles as regular posts. Describing a picture in the visible part of a post when the post isn't about the picture is horribly bad style. Doing so when there are multiple pictures in one post, regardless of whether Mastodon puts them in the right places (which it doesn't), is even worse.
I usually post pictures taken in #VirtualWorlds. In comparison with pictures taken in real-life, they have a much higher tendency to contain things that need to be described, often to both sighted and blind or visually-impaired users, because they simply don't know them, be it objects, be it locations. It's one thing if a picture was taken on Times Square, and it's something else if a picture was taken in a place of which maybe not even five people in the whole Fediverse even know that it exists. Thus, more text is needed.

Now there are two schools of thoughts when it comes to alt-text.

One: clear and concise alt-text. Only describe what's necessary in the context in which the picture is posted. Screen readers can't handle long alt-texts well. You can't navigate alt-text with most screen readers, i.e. you can't stop it somewhere, rewind it to a certain point and listen to parts of it once more. All you can do is let the screen reader rattle down the whole alt-text in one chunk. If you need to hear it again, you have to hear all of it again.

The obvious downside of this is that most of the content of the image is lost to everyone who isn't sighted, and some is lost to those who can't identify it even by looking at it in that particular picture.

Two: full description of absolutely everything in the picture plus explanation if necessary. Denying non-sighted people the chance to experience everything that's in a picture, and be it through words, can be considered ableist. Also, tiny details that are barely visible in the picture could be described so that sighted people can identify them.

And besides, there's the idea that alt-text can help everyone understand what that is that they see (or don't see) in that picture if they're unfamiliar with them.

As I've said, extensive image descriptions in the visible part of a post may be okay when the post is about the picture, but not when the picture illustrates the post and even less when there's more than one picture illustrating the post.

Yes, this is a thing. Just read what @Stormgren wrote earlier this month.

<img alt="Stormgren" src="https://hub.netzgemeinde.eu/photo/6994864a-99ce-4db2-974d-0733299644ef-6" height="32" width="32" />Stormgren wrote the following post Mon, 03 Jul 2023 18:20:44 +0200

Alt-text doesn't just mean accessibility in terms of low -vision or no-vision end users.

Done right also means accessibility for people who might not know much about your image's subject matter either.

This is especially true for technical topic photos. By accurately describing what's in the picture, you give context to non-technical viewers, or newbies, as to exactly what they're looking at, and even describe how it works or why it matters.

#AltText is not just an alternate description to a visual medium, it's an enhancement for everyone if you do it right.

(So I can't find any prior post of mine on this, so if I've actually made this point before, well, you got to hear a version of it again.)

And I'm actually waiting for Mastodon users to refuse to boost posts that contain pictures with insufficient alt-text. Many refuse to boost posts that contain pictures without alt-text already now.

The obvious downside of it is: "DESCRIBE ALL THE THINGS" + lots and lots and lots of stuff in the picture + just about everything needs to be explained because nobody is familiar with any of it = alt-text the size of a rather long blog post.

I've tried that with this picture (no embedding although I could because reasons). I've written a detailed alt-text. I've spent more than three hours in-world in a preserved, static copy of this place, researching and transcribing text where probably none of you would even know that there's text otherwise. The picture alone wasn't enough of a source for an alt-text that I would have deemed sufficient.

Only description plus some transcriptions: 7,636 characters. Description plus everything transcribed, save for the big black panel in the middle background behind the tree which I couldn't transcribe because it no longer exists in-world, plus translations of everything that isn't English plus everything unfamiliar explained: 10,985 characters. If that panel had still existed in-world, and I could have transcribed it, I might have passed the 12,000-character mark. With an image description.

As I've said, Hubzilla doesn't have a hard cap for alt-text length. In theory, it could handle and probably display alt-texts much longer than this. I don't know how it'd display an alt-text of that size in practice, whether it'd be scrollable, whether it'd have a time-out before anyone could read it fully etc. Mastodon, in the meantime, has the hard cap I've mentioned above which probably also cuts alt-texts coming in from outside. That's where most of my audience is. And screen reader users might have no other choice than to sit through their screen readers rambling down alt-text for more than five minutes in one go, especially if they could get a hold of the original alt-text instead of one cropped at the 1,500-character mark.

Now, even though I'll probably kick off two separate threads, I'd like to read your thoughts about how detailed alt-text should be.

#Accessibility #A11y #Inclusion #Inclusivity #InclusionMatters

19 comments

Looks like a question for [email protected]
- @WhoRoger If it were discussed only there, I guess it'd be rather one-sided:
  
  "450 characters tops, no matter what and how. No more than a few seconds for the screen reader. If the image is too complex, take simpler pictures."
  
  Then that's probably the answer you're looking for. Who else is gonna give you a better answer?
  
  I also think that makes quite sense. People with regular vision can just glance at a picture for half a second to get the idea of it, and move on. Having to spend a minute on every illustrative pic, no matter how unimportant, sounds more annoying than missing out on an extra layer of information. (Most of the time.)
  
  I used to be a journalist and I've learned you can trim a helluva lot of information without really losing much. It's a generally useful skill since time is limited for all of us.
  
  Ed: Also, for uses other than vision impairment, I think text should be elsewhere than alt-text. Like just description text or image metadata. Alt-text is for when you can't see the picture.
@jupiter_rowland @fediverse

tl;dr: Clear and concise alt-text goes in the alt attribute. Full description of absolutely everything in the picture plus explanation if necessary goes on the linked page with the link being placed immediately following the text.

Added: Be sure that link is meaningful, for example, "Alt text for image of the lighthouse control room" and not "Alt text for above image" since blind people will sometimes listen to a list of link text with no other context.
tl;dr?
@jupiter_rowland @fediverse I've wrestled with this over the years. Pictures of text is clear: it must contain all of the text. But in the situations you're describing, I'd say do enough to let the listener know if they want to learn more, then add a link to an alt-text blog post which gives the whole shebang. (I've heard varying reports that longdesc is inconsistently supported, so a separate link immediately below the image is best.)
- @jupiter_rowland @fediverse
  
  As for stuff that would be beneficial to non-technical viewers, that belongs in the separate alt-text blog post, since most folks not using a screen reader won't get to see the alt text attribute contents.
  
  This is all my take as a sighted reader who needs to prepare and audit website content and who rarely uses a screen reader.
  
  @jupiter_rowland @fediverse Also, my understanding of WCAG 2 is that it requires a text equivalent be provided, not that the text equivalent always be in an alt attribute.
  
  In particular, if you are going to provide an image of a table (something I'm appalled by, but it happens), that definitely belongs in the linked page, not in alt text, so that blind people can navigate the table to their preference rather than experience it as a data dump.
  
  #accessibility
- @ChasMusic (he/him) I could do an #AltText blog post, and I wouldn't even need anything beyond Hubzilla for it.
  
  However, I couldn't link to it below the image. That is, I could, at least here on #Hubzilla and for users of several other projects. But for the huge majority of #Mastodon users, the link would be entirely elsewhere because Mastodon can't display anything below images, save for other images that actually belong further up the same post. So this method is out of question for any posts with more than one picture because I can't draw connections between links in and pictures below the text that Mastodon users could see.
  
  And besides, I've got the feeling that many Mastodon users avoid links like the plague. They're using mobile phones, they practically never use the Web browsers on those things, and the Web browser popping up is an utter nuisance to them.
  
  Last but not least, I'm pretty sure that many Mastodon users are opposed to the idea of having the description of an image in a wholly different place that even requires a different app to view, rather than having it readily and neatly available right before them in their Mastodon app.
  
  @jupiter_rowland ¡Ouch! Okay, so working within those limits, you could:
  
  Toot the first toot with the image and brief alt text.
  
  Toot a second toot with just the full image description.
  
  Edit the first toot to provide a link to the second toot.
  
  The issue about link text is not, to my knowledge, solvable within current Mastodon. I need to file an issue if one has not already been filed. Obviously, you can't do anything which is not supported by the Mastodon code.

19 comments