You should learn how to properly write accessible image captions as the fediverse has a large amount of blind and visually impaired users who rely on screen readers

gabe [he/him]@literature.cafe to You Should Know@lemmy.world – 540 points – A guide to writing alt text and accessible image captions
mashable.com

Right now there is a bit left to be desired when it comes to lemmys accessibility features, but it's a good idea to be mindful of the fact the fediverse and its platforms tends to have pretty universal accessibility features that will likely come to lemmy sooner rather than later

41

I'll try to do this, but man it'd be great if there was an AI program that could auto caption/describe pictures as I post them. Or maybe just one that could interpret and describe everything on the screen, for a visually impared user.

There are some apps for Mastodon that do just that and hook into the alt text button that exists within Mastodon.

Alt-text descriptions should clearly convey both the content and the meaning of the image, and should aim to use as few words as needed. Describe what's essential to understanding (and enjoying!) the intent of the posted photo — you don't need to add in a sentence for every visual element, but should include as much as you need to create an accurate portrayal of the image. Cut out unnecessary words and combine separate sentences as much as possible. One to two sentences is usually more than enough room to describe what's going on.

As mentioned before, these photos convey information to the people scrolling your page, even if you are just posting them to brighten up your feed. They have a purpose, and for that reason, alt text should focus more on the image's meaning than its aesthetics. This means you're not focused only on what the object in the photo looks like, but what it is and why it was posted.

I was hoping to see a format that people can easily follow and just fill in the blanks, but I suppose this is the gist of it: Describe the main purpose of the photo succinctly rather than each and every individual thing you can see.

How about this?

Is the image purely decorative/e.g. if the picture never loaded does the audience miss out on some piece of information? If not then empty alt text suffices (this is less applicable to social media posts)

Otherwise imagine you have this picture and you’re telling somebody on the phone about it.

You’re not going to go into a full description of every element

“a small child on the left side of the picture is picking up a maple leaf with their index and thumb and handing it to the person on their right” who is a 50-60 year old woman with dirty strawberry hair and wearing horn rimmed glasses”

That’s too much and doesn’t “convey” the photo.

And you wouldn’t say “kid hands something to another person” since that’s not enough.

Let’s say you have an article about the fall and all the fun activities families can partake in. What matters is that it’s a young kid showing their grandma a leaf.

So something like this “Young child and grandmother at City Park. He is handing her a newly fallen leaf”

Or, if it’s an article about children nature education programs then it may say

“Teacher and a student in the park. The student is presenting a maple leaf to the teacher”

If it’s a post about your kid it may say

“Timmy showing his grandma a leaf he found while playing at City Park”

I can't really think of a case where a post's image isn't relevant to the associated discussion. Even if it's just a nonsense post with an unrelated image, knowing what the irrelevant image is about would help make sense of the discussion.

Right. If it’s a post it probably won’t be relevant.

Unless it’s something like

[birthday_cake.png] Birthday Party for Suzanne today[party_hat.png]

[suzanne.png]

Where: Breakroom Time: 2:00 PM

In that case I would add a blank alt text (so accessibility tools won’t read the file name) for the cake and party hat. Assuming it’s a platform that lets you add alt text per image.

But I would add alt text describing the photo of Suzanne “Suzanne answers the phone while accepting paperwork from a coworker”

You can edit it in after posting right? would be nice if people just went around on non-captioned posts and suggested text that the OP can simply paste in

Or somebody should build AI-powered image interpretation into screen readers.

Unfortunately those aren't very reliable, as our current iteration of AI is not very reliable. Most models to use from perpetuate a multitude of different bias pretty heavily.

And also, don't they need quite the processing power to be useful? Anybody know?

I know there was a subreddit or something that explained how to format the description from the context of Reddit. It would be cool if someone came up with a format to use as the standard for this here.

[This comment has been deleted by an automated system]

This is only applicable to inline pictures on text fields. Posting images/videos does not have an alt text field. The closest you can get is using the text field of a post as an alt text field.

The problem with this is images are thumbnailed but descriptions are not, so the first thing I see is the description and not the image, kind of ruining the impact of the image...if we could get all text descriptions hidden behind spoiler tags or something, that would be great

Yeah, I wish they would do it the same way Mastodon does. There's an "alt text" button for images

23 more...

Ok, how do we describe maymays then? Photo_of_man_bent_over_with_fingers_stretching_his....

Link to the knowyourmeme.com page, maybe?

Alternatively, for a lot of meme formats it would be appropriate to use the text embedded in the image as the alt text.

I tried writing alt text with the image on this post on August 9, and I thought it wasn't working. On mine and other Kbin instances, alt text does load but Firefox just doesn't show it when I mouseover. In the page's source, it's an "alt" argument instead of a "title" argument, and Firefox users can't see alt text if the image loads successfully. It would work better if Kbin allowed users to input title text instead of alt text.

I didn't notice until just now, but that post's alt text works perfectly on Mastodon! It simply duplicates the alt text into the title text and I can read it when I mouseover the image.

That leaves Lemmy which seems to have no alt text support at all. I've inspected the page and my alt text simply doesn't federate to Lemmy. But Lemmy apparently has working spoiler tags and Lemmy users are considering using that to describe images. The Lemmy spoiler tags don't work anywhere else, though.

::: spoiler Lemmy spoiler tag test
Everywhere but Lemmy will just render this as regular text. My Wordle-playing friends would be sad.
:::

I really like the text size options Voyager and Memmy.

Much better than text size options on apps for that other site.

Sometimes I think some additional attention to the formatting could be given, like if devs could test more with larger text size. But it’s still pretty good.