[quote]Well... I don't think anyone's necessarily expecting perfection. They just expect a reasonable and deliberate effort.[/quote]
They may not expect perfection all over. But they expect full accessibility for [i]themselves[/i].
[quote]There are some guidelines out there from WGBH and other sources that can help decide what is and isn't necessary to have a well-described image.[/quote]
Trust me, I know these guidelines. I follow them to a tee as far as that's possible for me. They have influenced me greatly and still do.
Just recently, I've learned two new things from such a guide. One, dimensions must always be given as relative to a size that people, especially blind or visually-impaired people, are familiar with, e.g. human body height. Two, colours must always be described based on a specific selection of basic colours plus brightness/darkness plus saturation. Blind people have no concept of "cerise" or "crimson" or "mocha". These two points rendered all my previous image descriptions obsolete.
But my images are extreme edge-cases that put these guidelines and the general image accessibility rules to the test.
Let me give you a few examples.
First of all, in order to keep an image description, especially in alt-text, short and concise, the guides tell me to focus on what's important in the image in the context of the post, describe that and sweep everything else under the rug because it's unimportant.
This advice completely failed [url=https://hub.netzgemeinde.eu/item/c8a14063-b4e2-48fa-baf9-cb4faef7225b]here[/url] because there are [i]several dozen[/i] singular elements in the image that are important within the context of the post. I did describe everything in the images, but only a tiny fraction of the 40,000-character description of the first image is about the surroundings.
It failed even more [url=https://hub.netzgemeinde.eu/item/136f021c-06cb-4e38-abe6-2bf37fd521d8]here[/url] and especially [url=https://hub.netzgemeinde.eu/item/8c2b4728-dda5-498b-9f84-2f11e163a4a5]here[/url]. In these cases, I wanted to show [i]the whole scenery with everything in it[/i]. Nothing was more or less important in the images within the context of the post than anything else. Thus, I had to describe everything because, in a sense, everything was important.
It also failed [url=https://hub.netzgemeinde.eu/item/f8ac991d-b64b-4290-be69-28feb51ba2a7]in this image post which I've already linked to earlier in the thread[/url]. Again, I wanted to show the whole scenery. But even if I had wanted to show only the central element of the image, the building, I would not have been able to shrink the image description to a sensible size. The building is so complex and so strange and unusual at the same time that a highly detailed and very long description of it was absolutely necessary and inevitable. It's for this reason why I normally try to avoid having buildings in my images.
Another advice is to mind the target audience and write for them. This advice always comes in guidelines for alt-texts on static Web sites or blogs. It never comes in guidelines for alt-texts in social media.
Static Web sites or blogs have very defined target audiences, namely those who actively seek out the kind of content that is on these sites and comes from these blogs. Especially if they're about science, technology or the like, the Web designer or blogger can always count on the target audience being interested in the field and, thus, coming with substantial prior knowledge. Certain things simply don't have to be explained to this target audience because they already know them.
The target audience on social media, on the other hand, can be [i]anyone[/i]. In the Fediverse, my target audience includes people who happen upon my posts on the federated timelines of their instances.
People who have one of my image posts boosted/reposted/renoted/repeated to them for whatever reason.
People who discover one of my image posts on their local timeline because someone on their instance has boosted it.
People who come across one of my image posts among the posts of someone whose profile they're checking out, but that user has only just boosted one of my image posts. Granted, I don't think anyone has ever boosted any one of my image posts because they're too intimidating for Mastodon, especially not @[url=https://mastodon.social/@alttexthalloffame]Alt Text Hall of Fame[/url], but it could always happen.
And, of course, people who happen to follow one of the hashtags I use, and that hashtag is not one about the topic (e.g. [code]#[zrl=https://hub.netzgemeinde.eu/search?tag=OpenSim]OpenSim[/zrl][/code]), but it happens to be one that I use as a filter trigger (e.g. [code]#[zrl=https://hub.netzgemeinde.eu/search?tag=LongPost]LongPost[/zrl][/code]). Or they're interested in all things alt-text, so they follow [code]#[zrl=https://hub.netzgemeinde.eu/search?tag=AltText]AltText[/zrl][/code], but I put that hashtag on my image posts as well.
Now, my image posts are about a general topic with which maybe one in over 200,000 Fedizens is familiar. The chances that someone who happens upon one of my image posts is familiar with the general topic are practically nil.
Thus, taking my target audience into account, I also have to take into account that at least parts of my target audience know [i]nothing[/i] about what my image posts are about. Absolutely zilch. Not even the basics.
For starters, this means I have to explain everything from the ground up. Look at the image posts I've linked to again. Open the content warnings again. Check the beginnings of the full image descriptions. Notice how much I have to ramble in order to only explain where the image was taken.
In real-life pictures, you'll probably never need more than 70 characters to name even the most obscure real-life location. In fact, you'll often get away with name-dropping because most people are familiar with the place or even a specific landmark anyway.
In my virtual world pictures, I can't get away with name-dropping. I have to go all the way and explain on which sim a landmark is, in which grid the sim is, that the grid is a virtual world based on OpenSim, what OpenSim is, what Second Life is, what grids are, what this particular grid is, why they're named grids, what regions are, what sims are, even what the Hypergrid is in order to make clear how my avatar got to that particular location in the first place. None of this can be expected to be known by everyone in the target audience.
Mentioning the place where I've created a picture often requires more characters than fit into Mastodon's alt text.
Likewise, most people should at least have a rough idea what the Capitol is and what it looks like. Or the Eiffel Tower. Or the Taj Mahal. Or a real-life cat, apart from the ways that cats look different from one another.
How many people, do you think, know right off the bat what the Sendalonde Community Library is and what it looks like? Would you know without Googling?
Or how many people, do you think, know right off the bat what an OpenSimWorld beacon is and what it looks like? You think I could get away with name-dropping "OpenSimWorld beacon" in an image description? It doesn't help that there are at least five standard variants from various versions, and some sim users customise them or build their own.
Or if I just simply wrote that there's a teleporter somewhere, how many people, do you think, would be sufficiently informed? How many non-sighted people would be sufficiently informed?
I could post a picture that includes a virtual version of the Eiffel Tower or the Golden Gate Bridge. Really, I could, I know where to find them. Blind or visually-impaired people may have a rough idea what they look like in the real world. People with bad mobile Internet for whom my image doesn't load may know exactly what these structures look like in the real world. But how are they supposed to know what the virtual models of these structures look like? And how is anyone supposed to know where they are located?
So, yes, I do take into account that parts of my target audience may not be fully sighted. That's why I'm doing all that in the first place. That's why I describe stuff that many people may deem unimportant.
See, I'm posting about 3-D virtual worlds. For many people, that's completely uninteresting. And I guess not few actually filter the word "metaverse" out for whatever reason.
But there may be people who have actually put high hopes into virtual worlds without knowing that virtual worlds already existed at that point. Then they heard about "the metaverse" dwindling away. And they were disappointed.
And then they come across one of my image posts. About a 3-D virtual world. And they're immediately on the edges of their seats in utter excitement. There are actually [i]existing[/i] virtual worlds? Like, [i]right now[/i]? That picture in that post is not an AI rendering, but it was [i]made in an actual, real, living 3-D virtual world[/i]?!
If they read on, they'll discover on top of all that that the virtual world which my image shows is free and open-source. And decentralised. You know, like Mastodon. To the point where anyone can run their own instance. Or rather, grid. And connect it to all the other grids.
This is not what they've dreamed of. This [i]exceeds[/i] what they've dreamed of. They would never even have dared to [i]dream[/i] of something like this because it was so very unimaginable. And now it turns out it's very real. Oh, and it has been for 17 years already.
Now they're curious like you wouldn't believe. They want to know [i]everything[/i] about these worlds. They want to soak up any information about them they can get like a sponge. They want to explore these new and unknown worlds.
If they're sighted, they start by letting their eyes wander around the image. No matter what the image focuses on within the context of the post, they check out [i]everything[/i], also because all that stuff looks much, much more realistic and much, much less cartoonish than those official previews from Meta's Horizons which all of a sudden look even more ridiculous and pathetic in comparison.
I could post a virtual cat picture. No, really, I could. They wouldn't just look at the cat and go, "Aww, how cute." They would check out the surroundings and the background just as well, even though the surroundings and the background don't really matter. But to them, they [i]do[/i] matter because it's such an exciting new world.
Now, if these people were blind or visually-impaired, one might argue that they wouldn't be nearly as interested in such a highly visual medium as a 3-D virtual world. I wouldn't count on that. Blind or visually-impaired people might be just as curious about these new worlds as sighted people and just as eager to go explore these new worlds. They might be just as eager to know what [i]everything[/i] in [i]any[/i] of my pictures looks like, regardless of context. But they need my help.
In general, I don't buy the notion that blind or visually-impaired people can easily be satisfied by mentioning stuff that's in an image, that you can get away with mentioning and name-dropping and nothing more.
I always expect blind or visually-impaired people to be all like, "Yeah, that's fine, but [i]what does it look like[/i]?"
The more curious they are about something, the more eager they are to know what it looks like. And the newer and/or more obscure something is to them, the less likely they are to already know what it looks like. And the virtual worlds I write about are [i]completely unknown[/i] to them at this point.
All this ties in with the next point. And that's to avoid technical terminology and jargon. If you can't, you have to explain it in such a way that even the most casual audience can fully grasp it with no problems and without having to look up anything themselves. Anything less is ableist.
I can't fully avoid it. If I tried, I would have to constantly write around it in such a way that my image descriptions would get even more confusing.
So I have to explain it. All of it. Down to the very basics. Down to something that I can expect everyone in the Fediverse to know and be familiar with.
Seeing as how very niche and obscure and special-interest my virtual worlds topic is, and how many unfamiliar elements will inevitably appear in both my images and their descriptions, I have a whole lot to explain. Even explaining Second Life would be easier than explaining OpenSim because explaining OpenSim [i]requires[/i] a previous explanation of Second Life.
Anyone who believes that I can cover all that in significantly under 500 characters in a way that absolutely everyone can understand right away with no special prior knowledge, without Googling, without asking me questions afterwards, I have a bridge to sell you.
Then there is how people should be described. I only post photographs of real-life people if they're part of a known, established meme template.
Otherwise, virtual avatars take their place. I have started to avoid posting avatars after I had learned about the eye contact trigger. The issue here is two-fold: Mastodon doesn't hide images behind content warnings. And Hubzilla can't make Mastodon blank out sensitive images. So any sensitive image which I post on Hubzilla will always end up immediately in plain sight for everyone, at least for those who don't have filters that catch these image posts.
When I couldn't really avoid having an avatar in an image, I showed the avatar from behind so that the face was completely invisible. It still substantially increased the effort of description and the length of the description, but it was mostly safe from triggering people, and I didn't have to go through the effort of describing the avatar's face and mimics.
Now I'm also on (streams) which can make Mastodon blank out sensitive images using one out of two specific hashtags. So there's no stopping me posting pictures with avatars in them and even showing their faces.
Describing avatars can end up tedious, though, much more tedious than describing comparable real-life humans, applying the same description rules.
First of all, there's no skimping. There's no "it doesn't matter what the avatar looks like". It [i]always[/i] matters to someone what the avatar looks like. Again, exciting new world to explore and all. See above. And if the image is a portrait of the avatar, it matters [i]very much[/i] what the avatar looks like.
Next, there's much less that doesn't have to be explained because it's a given. For example, there's quite a bit of debate on whether or not to mention a real-life human's skin tone, and if so, how. The general consensus is usually to limit it to "light, medium light, medium, medium dark, dark". Nothing much more is necessary except maybe freckles or something. But a real-life human skin always looks like a real-life human skin. That's a given, isn't it?
In a virtual world, it isn't a given. In case you were unaware: The skin of a human avatar in a virtual world can be a plain, solid tint on a 3-D mesh. It can be photo-realistic. It can be one or multiple actual photographs, e.g. because the whole avatar was generated from photographs and a 3-D laser scan of a real-life human. And it can be anything in-between.
Look up early promo pictures showing Mark Zuckerberg's personal Horizons avatar. Then search Flickr for Second Life pictures. There are [i]worlds[/i] between these two. You may be tempted that the Second Life pictures were generated by Stable Diffusion rather than rendered in a virtual world.
So if I just mentioned the skin tone, how would you, without looking at the image, be able to know what exactly the skin looks like, whether it's a cartoonish solid colour or a set of photo-realistic textures?
This goes on and on.
In fact, when I describe an avatar, I also have to [i]explain[/i] the avatar. With that, I mean I have to explain the avatar system in Second Life and OpenSim and how it works. Unlike in many other virtual worlds, they aren't single-piece "monoblock avatars" Γ la Ready Player Me that can only be varied in overall size, if at all. They are highly customisable in-world with no external tools, and they are highly modular. This explains why avatars looks like what they look like. And seriously, only experienced Second Life and/or OpenSim users even expect this level of in-world customisability and modularity to be possible in the first place.
If I were to go all the way, I'd even explain what mesh is, what rigged and fitted mesh is, how it works, how it differs from the "classic" means of customising Second Life and OpenSim avatars, what Bakes-on-Mesh is and what advantages Bakes-on-Mesh has over what was used before it was there.
In [url=z6Mkmc3YmgUu5jTyhc6YqC8VjnMwmFtdjFFA45MHTqyBFaA2/item/a1d77470-fd61-403c-8472-74760995bb91">https://streams.elsmussols.net/display/?mid=https://streams.elsmussols.net/.well-known/apgateway/did
z6Mkmc3YmgUu5jTyhc6YqC8VjnMwmFtdjFFA45MHTqyBFaA2/item/a1d77470-fd61-403c-8472-74760995bb91]my latest virtual world image post[/url], I omitted the latter, and I also forgot to mention that the skin textures are near-photo-realistic, and the skirt texture is photo-realistic. Maybe I'll edit the image description preamble in the post and add both.
But what increased the effort of describing the avatar are the OSgrid logo and the logo for OSgrid's 17th birthday. Nobody who isn't an active OpenSim user knows what either looks like without seeing it. So I had to describe both in detail, also because the latter includes three independent lines of text which I had to transcribe.
This brings me to the last point: text in images and its transcription.
There is the rule that says that any and all text within the borders of an image must be transcribed 100% verbatim.
In the Fediverse, this rule is usually applied to π screenshots, Mastodon screenshots, pictures of newspaper clippings or photographs of posters. This is usually manageable except when the text in a newspaper clipping is over 1,500 characters long.
The various alt-text guides out there that mention text transcripts are never about social media, what with how the puny character limits for alt-text in commercial social media makes transcribing text difficult to nigh-impossible.
They're always about static Web sites and blogs. This mostly means commercial Web sites, professional scientific Web sites, professional scientific news blogs or professional Websites or blogs about technology.
So some of the guides know the occasional edge-case. A graph or a flowchart is a near-edge-case due to its complexity. There are specific defined rules on how to describe graphs, although they may differ. One solution is to fully describe a flowchart in the main text. As for graphs, it's often considered okay to sweep the exact details under the rug unless every last figure in the graph absolutely matters. A complex table is very much an edge-case because a full transcript of it in alt-text would be too long. It is not as clear how exactly a table has to be handled, but it is being discussed.
Most importantly, both of these edge-cases are very well-known. They happen to professional Web designers and bloggers over and over and over again, and they have been happening for many, many years.
My images are edge-cases, too. But they're completely unknown edge-cases, and they're edge-cases in ways that are completely unknown to all those who write alt-text guides and who work on standardising accessibility.
I often have lots of bits of text in them in various places and ways. Signs, posters, box labels, vendor labels, teleporter labels, flags, neon signs, prints on clothing, the writing on an OpenSimWorld beacon, what-have-you. All this could be in my images. And a lot of it has already been in my images, sometimes in great numbers.
This alone is an edge-case. An edge-case which none of the alt-text guides handle because none of them are even aware of its very existence. The same situation could occur in real-life photographs, but real-life photographs practically always focus on one specific element or a few, and everything else is recommended to be treated as non-existent in alt-text. So part of my edge-case is that this focus on one specific text-less element is simply not there in most of my images.
I can solve it gracefully by [i]actually transcribing absolutely everything verbatim, 1:1, 100%[/i]. It usually doesn't go into the alt-text anyway unless it isn't too much. It definitely always goes into the long, full, detailed description in the post, in case you've forgotten that I usually describe each one of my images twice. Only if it's a maximum of three or four short bits of text, and I don't have to describe too much in the alt-text otherwise, only then the transcripts go into the alt-text as well. But I never put some transcripts into the alt-text and some not. Either all of them or none. Regardless, all of them always go into the full description in the post.
What makes my unhandled edge-case even more of an unhandled edge-case is that text often tends to be illegible in my images.
It may be too small to be readable at the resolution at which I post the image, e.g. two or three pixels high. It may be so small that it can't be identified as text at that resolution. It may be so tiny that it's practically invisible, e.g. less than half a pixel high. But still, it's text, and technically, it's within the borders of the image. Or it could be partially obscured, e.g. a poster with a lot of text of it, but 30% of the front area of the poster are covered by the trunk of a tree in front of it.
No alt-text or image description guide that deals with text transcripts [i]ever[/i] deals with unreadable text. They don't declare that it must not be transcribed. They don't declare that it needn't be transcribed. They don't declare explicitly that it must be transcribed.
The very existence of illegible text in an image completely eludes absolutely everyone who writes guidelines for image descriptions.
Given the lack of special treatment requirements of illegible text, I have to suppose that illegible text has to be handled the exact same ways as legible text. It has to be transcribed verbatim. And so I transcribe it verbatim.
Now you may say that that shouldn't even be possible. If that text is illegible, I can't read it either, so how should I be able to transcribe it?
But I [i]can[/i] read it.
See, when I describe an image, I don't just simply look at the image at the resolution at which I intend to post it. I look at [i]the real deal[/i] in-world. I don't look at the representation of a sign in my image. [i]I look at the sign proper[/i], in the virtual world, right where the sign is installed.
I can walk around with my avatar. I can move the camera independently from the avatar. I can move around obstacles such as tree trunks. I can close up on everything.
And all of a sudden, that tiny white blot of 4x3 pixels on which the writing can't even be made out as such unfolds before my eyes as a sign with a front texture of 1024x768 pixels with half a dozen clearly readable lines of text.
And all of a sudden, I can read that text behind that tree.
In fact, I'd say that even [i]fully sighted[/i] people can profit from how I transcribe everything, legible or not. If even the most eagle-eyed Fediverse users can't read something in one of my images, they still know what's written there because I can read it, and I tell them what's written there.
Lastly, all this is just me following existing guidelines and rules, from explaining jargon and unknown words to describing people to transcribing text to taking my audience into account and giving them what they require. It's just that my images are extremely obscure edge-cases for which no officially defined and agreed-upon exceptions exist and probably never will.
#[zrl=https://hub.netzgemeinde.eu/search?tag=Long]Long[/zrl] #[zrl=https://hub.netzgemeinde.eu/search?tag=LongPost]LongPost[/zrl] #[zrl=https://hub.netzgemeinde.eu/search?tag=CWLong]CWLong[/zrl] #[zrl=https://hub.netzgemeinde.eu/search?tag=CWLongPost]CWLongPost[/zrl] #[zrl=https://hub.netzgemeinde.eu/search?tag=OpenSim]OpenSim[/zrl] #[zrl=https://hub.netzgemeinde.eu/search?tag=OpenSimulator]OpenSimulator[/zrl] #[zrl=https://hub.netzgemeinde.eu/search?tag=Metaverse]Metaverse[/zrl] #[zrl=https://hub.netzgemeinde.eu/search?tag=VirtualWorlds]VirtualWorlds[/zrl] #[zrl=https://hub.netzgemeinde.eu/search?tag=AltText]AltText[/zrl] #[zrl=https://hub.netzgemeinde.eu/search?tag=AltTextMeta]AltTextMeta[/zrl] #[zrl=https://hub.netzgemeinde.eu/search?tag=CWAltTextMeta]CWAltTextMeta[/zrl] #[zrl=https://hub.netzgemeinde.eu/search?tag=ImageDescription]ImageDescription[/zrl] #[zrl=https://hub.netzgemeinde.eu/search?tag=ImageDescriptions]ImageDescriptions[/zrl] #[zrl=https://hub.netzgemeinde.eu/search?tag=ImageDescriptionMeta]ImageDescriptionMeta[/zrl] #[zrl=https://hub.netzgemeinde.eu/search?tag=CWImageDescriptionMeta]CWImageDescriptionMeta[/zrl] #[zrl=https://hub.netzgemeinde.eu/search?tag=Transcript]Transcript[/zrl] #[zrl=https://hub.netzgemeinde.eu/search?tag=Transcripts]Transcripts[/zrl] #[zrl=https://hub.netzgemeinde.eu/search?tag=Transcription]Transcription[/zrl] #[zrl=https://hub.netzgemeinde.eu/search?tag=Inclusion]Inclusion[/zrl] #[zrl=https://hub.netzgemeinde.eu/search?tag=Inclusivity]Inclusivity[/zrl] #[zrl=https://hub.netzgemeinde.eu/search?tag=A11y]A11y[/zrl] #[zrl=https://hub.netzgemeinde.eu/search?tag=Accessibility]Accessibility[/zrl] #[zrl=https://hub.netzgemeinde.eu/search?tag=AccessibilityMatters]AccessibilityMatters[/zrl]