@Alt Text Hall of Fame In case you've just heard about it for the first time:

Here it is (and still a WIP).

#FediMeta #FediverseMeta #CWFediMeta #CWFediverseMeta #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta #Wiki

@Alt Text Hall of Fame I really need to work more on my alt-text and image description wiki.

I don't know how many users will find a wiki with 50+ pages useful, but all that information must be gathered in one place, adapted to the culture and technology of the Fediverse and especially to Mastodon's culture and published for people to read.

I mean, the "how" has to include the elimination of quite a number of mistakes that just about everyone in the Fediverse keeps on making because they don't know that it's wrong.

#Long #LongPost #CWLong #CWLongPost #FediMeta #FediverseMeta #CWFediMeta #CWFediverseMeta #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta

@Mrs. McGibbons 🧚‍♀️ This may be true for real-life cat photos. Just about the most simple images imaginable in the Fediverse.

But what if I don't just post real-life cat photos in which I don't have to describe much more than the cat? Because I don't post real-life cat photos. I don't post real-life photos at all unless they're meme templates.

My own original images are renderings from 3-D virtual worlds. Very obscure 3-D virtual worlds even. This means:

• I can't forgo detail descriptions under the assumption that people know what stuff looks like anyway. I can't assume that anyone already knows what anything in my images looks like.
• Every other time, there's nothing particular in the image that matters more within the context of the post than everything else. Instead, the whole image with everything in it matters all the same.
• The other half of times, what matters within the context is irrelevant because the existence of 3-D virtual worlds is so intriguing to someone out there that they demand a full, detailed image description.

As for the actual alt-texts, I'll try to keep them at 512 characters or fewer, difficult as that will be. But I'll do that for technical reasons: While Misskey and its forks are supposed to truncate longer alt-texts at the 512-character mark, they actually delete them due to a bug. If I make them longer, users on Misskey, Sharkey, Iceshrimp-JS etc. will believe that I haven't written any alt-text in the first place.

But I will keep adding long, fully detailed image descriptions to the post text where I have much more room. I need room for sufficiently detailed descriptions, I need room for all the explanations necessary for people to understand the post and the images and the descriptions, and I often need room for all the text transcripts.

For example, do you know what the main building on Nebadon Izumi's Universal Campus looks like? Would it be sufficient for you if I just name-dropped it? Or must I describe what it looks like?

If so, well, that's 40,000 characters of description only for that building and what's visible inside it because the building mostly has glass panes for walls. 7,800 characters only for the front of a building that's five times as long as it's wide. 500 characters only for that one piece of structure around the main entrance doors. In fact, over 1,600 characters for the doors. Also, 3,200 characters for a teleport panel, including transcripts of 13 bits of text. Been there, done that, got the figures from there.

Don't worry, I will always hide long posts behind a summary with content warnings, including a warning about the post being tens of thousands of characters long due to the long image descriptions.

In fact, my meme posts will continue to be very long themselves, although not quite as long as posts with original pictures. Describing the visuals is easy most of the time, and it can be done in 512 characters or fewer. But they still need explanations. Otherwise, nobody will understand anything. All my meme posts are about obscure topics, too.

Now I'm wondering what's more likely to upset people and make them sanction me in some way, including blocking me without saying a word. Insufficient image descriptions? Insufficient alt-text in particular? Not putting all the text transcripts into the alt-text where many insist that they belong? Or posts behind summaries and CWs that indicate that these posts are 25,000, 40,000, 60,000 characters long?

But seriously, even if I cut down visual descriptions to a more normal level, which would come with its own nasty side-effects, I would still need to explain everything. So no, I can't keep image posts at 500 characters or fewer.

#Long #LongPost #CWLong #CWLongPost #FediMeta #FediverseMeta #CWFediMeta #CWFediverseMeta #CW #CWs #CWMeta #ContentWarning #ContentWarnings #ContentWarningMeta #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta

@europlus :autisminf: @R.L. Dane :Debian: :OpenBSD: :FreeBSD: 🍵 :MiraLovesYou: Although, honestly, there are quite a few things I'd criticise about the alt-texts.

#AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta

@Cassandrich @Sobri | Zoe (she/her) @Scott Jenson @Phil Dennis-Jordan Also, an image doesn't always need the exact same alt-text whenever it's posted somewhere.

The alt-text must adapt to the context. It must be different according to the context in which an image is posted. Also, it must adapt to the place where it's posted. The same image, even within a very similar context, must have a different alt-text in the Fediverse than on commercial social media or a static website. Lastly, and this ties in with the Fediverse requiring different alt-texts, the audience must be taken into consideration.

Alt-text in metadata can't do either of this. An LLM can't do either of this either unless it's explicitly prompted to do so, and even that is questionable.

Many Mastodon users dream of only pressing a button or not even that, and some AI automagically generates a perfect alt-text for their image. Perfectly accurate with exactly the details required for the context and the intended audience as well as the expected audience, all while following every last image description and alt-text rule out there to a tee.

It's perfectly understandable. Mastodon had begun to feel like child's play when they were suddenly pressured into describing each and every image they post. Worse yet, it seems like over 90% of all Mastodon users do everything on a phone with no access to a hardware keyboard whatsoever. So they have to fumble their alt-texts into a screen keyboard while not even being able to see the image they're describing.

I'm neither on Mastodon nor on a phone. I've got the luxury of having a desktop computer with a hardware keyboard and being able to bllind-type. So I don't have a problem with writing my image descriptions myself with no help from an AI.

In fact, my own original images are all about an extreme niche topic. It's so obscure that no AI will ever be able to describe such images, much less explain them at my level of accuracy and detail. (Explanations go into the post text, by the way, and not into the alt-text, but I always have an additional image description in the post text for my original images anyway.)

I simply know things that no AI will ever know, not ChatGPT and not Claude either, at least not at the point in time when they need that knowledge. And I can see things that will always remain invisible for AIs.

You can develop better models all you want. But they'll never be able to do all that.

#Long #LongPost #CWLong #CWLongPost #FediMeta #FediverseMeta #CWFediMeta #CWFediverseMeta #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta #AI #AIVsHuman #HumanVsAI

@Ahimsa @Paul_IPv6 @Izzy I'm occasionally working on my own extensive wiki about alt-texts and image descriptions in the Fediverse. It's still very much a WIP, and not even half of the planned pages are done, and it specialises in the Fediverse (not only Mastodon, by the way). But maybe you'll find something there that's useful for static websites as well.

Here you go.

If that shouldn't suffice, I've got more than 50 articles, pages etc. about alt-texts and image descriptions linked on this page, including 25 articles by Veronica Lewis a.k.a. Veronica with Four Eyes.

#Long #LongPost #CWLong #CWLongPost #FediMeta #FediverseMeta #CWFediMeta #CWFediverseMeta #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta

@C. I have two major issues with the Mastodon HOA.

One, they try hard to force "Mastodon standards", Mastodon culture and Mastodon's unwritten rules upon the whole Fediverse. Including places that not only aren't Mastodon, but that are very much not Mastodon. Simply because they can't see where a message is from. In fact, many of them are still fully convinced that the Fediverse is only Mastodon.

And so you have members of the Mastodon HOA yelling at someone who is allegedly "doing Mastodon wrong", but that someone is actually on Friendica and has been since as early as 2011. As in about five years longer than Mastodon has even existed. And seriously, the only places in the Fediverse that are even more different and farther away from Mastodon than Friendica (without specialising in something that Mastodon absolutely can't do) are Friendica's own descendants: Hubzilla, (streams), Forte.

The Mastodon HOA probably don't know that Friendica exists. They definitely don't know that either of the other three exists. They definitely don't know that any of the four is significantly different from Mastodon in any way. And frankly, they don't care a bit. If it appears on any Mastodon timeline, it's Mastodon to them, and it has to adapt to Mastodon's culture and follow Mastodon's rules.

Two, they don't coordinate anything among each other. They're just a bunch of lone wolves. Everyone has got their own standards, but everyone thinks their personal standards are the one and only Mastodon/Fediverse gold standards, and everyone enforces their own standards. And, of course, everyone thinks their standards can and must apply always, including in the most obscure edge-cases.

For example, they've got standards for describing real-life photos on Mastodon with a character limit of 500. And they try to enforce these standards always and everywhere. However, these standards don't necessarily work perfectly when I post a rendering from a super-obscure 3-D virtual world on (streams) with a character limit of over 24 million where I've got loads of room to write an additional long image description and put it into the post text.

The Mastodon HOA, or at least some of their members, appear to be constantly raising their minimum quality requirements for image descriptions. They must be absolutely accurate, and they must be sufficiently detailed that nobody will ever have to ask for a detail description. Oh, and they must explain whatever the audience may not know about the image or the description. (At this point, it's fair to mention that explanations must never go into the alt-text.)

Sure, I can do that. I have done so in the past. But I can't do that within Mastodon's alt-text character limit of 1,500 (Mastodon truncates longer alt-texts from outside). I can do that even less within Misskey's alt-text character limit of only 512 (Misskey and the Forkeys should truncate longer alt-texts, but due to a bug, they delete them entirely instead, giving the impression that you haven't written an alt-text at all). I can only do that in the additional long description in the post text.

If the Mastodon HOA demand I transcribe literally any and all text within the borders of an image, I can do that, too. In fact, I have done so in the past. I can transcribe bits of text verbatim which the Mastodon HOA can't even read. Which the Mastodon HOA couldn't even find in the image because they're so tiny. But there's no way that I can squeeze 20+ individual text transcripts into 1,500 characters of alt-text along with the rest of the visual description, much less into only 512 characters. The text transcripts will have to go into the long description in the post text, whether the Mastodon HOA want or not.

This means that the post will exceed the holy limit of 500 characters by huge magnitudes. This, in turn, means that when I've satisfied one Mastodon HOA member, another one comes and sanctions me for exceeding the holy 500-character limit. That is, chances are it's actually the same Mastodon HOA member.

In other words, if the content of an image is obscure enough and requires enough description, the only winning move when I want to post such an image is to not post it at all.

#Long #LongPost #CWLong #CWLongPost #FediMeta #FediverseMeta #CWFediMeta #CWFediverseMeta #CharacterLimit #CharacterLimits #CharacterLimitMeta #CWCharacterLimitMeta #500Characters #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta #MastodonCulture #MastodonHOA

@Woochancho @Diego Martínez (Kaeza) 🇺🇾 @🅰🅻🅸🅲🅴  (🌈🦄) Especially whenever humans have advantages over LLMs.

When I describe my own original images, I have two advantages.

One, I know much more about the contents of the image than any AI. That's because my original images always show something from extremely obscure 3-D virtual worlds. On top of that, I may add some extra insider knowledge or explain pop-cultural references in the long description in the post if it helps understand the image and its descriptions.

Two, the LLM can only look at the image with its limited resolution. That's all it has. In contrast, when I describe my images, I don't just look at the images. I look at the real deal in-world with a nearly infinite resolution.

For example, an LLM can only generate a description from a picture of a virtual building. But when I describe it, my avatar is in-world, standing right in front of the building whose picture I'm describing. I can move the avatar around, I can move the camera around, I can zoom in on anything. I can correctly identify that four-pixel blob as a strawberry cocktail wheras the LLM doesn't even notice it's there.

I've actually done two tests using LLaVA. I've fed it two images I had described myself previously to see what happens. It was abysmal. LLaVA hallucinated, it interpreted stuff wrongly and so forth, not to mention that LLaVA's description, even after being prompted to write a detailed description, wasn't nearly as detailed as mine.

In one image, there's an OpenSimWorld beacon placed rather prominently in the scenery. LLaVA completely ignored it. I described what it looks like in about 1,000 characters, and then I explained what it is, what OpenSimWorld is and how it works in another 4,000 characters or so.

It's an illusion that AI will soon catch up with any of this.

Oh, by the way: How is an AI supposed to pinpoint exactly where an image was made if the image shows a place of which multiple absolutely identical copies exist? Or if the image has a neutral background that doesn't even hint at where it was made? I can do that with no problem because I remember where I've made the image.

#Long #LongPost #CWLong #CWLongPost #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta #AI #LLaVA #AIVsHuman #HumanVsAI

@Pino Carafa Well, my problem is not the alt-text.

I used to limit my alt-texts to 1,500 characters because Mastodon and its forks truncate longer alt-texts at the 1,500-character mark. In the future, I will limit them to 512 characters because Misskey and its forks should truncate them at that mark if they're longer, but instead, they delete them.

But in addition to my alt-texts, I describe my original images once more (= twice altogether). The other description is what I call the "long description", and it goes directly into the post text (as opposed to the alt-text). I don't have a character limit to worry about (over 16.7 million), so I can do what's outright unimaginable from a Mastodon point of view.

It's this long description that's causing trouble.

That is, I wouldn't wonder if the Mastodon HOA were to sanction me for my alt-text not being detailed enough when I limit it to 512 characters. In fact, I wouldn't wonder if they were to sanction me because a 1,500-character alt-text of mine is lacking important elements (descriptions of certain details, transcripts of all text within the borders of the image etc.).

#Long #LongPost #CWLong #CWLongPost #FediMeta #FediverseMeta #CWFediMeta #CWFediverseMeta #CharacterLimit #CharacterLimits #CharacterLimitMeta #CWCharacterLimitMeta #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta #MastodonHOA