Helping a friend with a buying decision, and remembered something that works for me when there are multiple options in the same category and you can't decide between them.
* Make a list of candidates. * Roll some dice to pick a candidate at random. * Investigate your feelings about the dice results. Are you happy the dice picked "the right one"? Or are you disappointed? If you are disappointed, strike that candidate from the list and roll again.
Rechazando cualquier regalo, favor o servicio en condiciones ventajosas que vaya más allá de los usos habituales, sociales y de cortesía, sin perjuicio de lo establecido en el Código Penal.
Hey @NVAccess. Are you aware the NVDA expert certification website appears to be down? Or has it moved? My browser throws the following message at me when I try to visit it. Looks like there’s a problem with this site certification.nvaccess.org/ might have a temporary problem or it could have moved. Error code: 502 Bad Gateway • The site could be temporarily unavailable or too busy. Try again in a few moments.
I must be old. I really don't understand emoji reactions. If I say "I'm on the bus and will be there in half an hour" to the group message, why do five people need to react with the bus emoji? A thumbs up or a heart or a smile I get. It communicates some sort of information. You like it, or you're happy, or whatever. But all reacting "bus" communicates is that...the word bus was in the original message, and you recognized it! Yes! Good for you! You recognized the word bus! Do you want a cookie or something?
@tewha So what you're saying is there's at least a couple guys in the group message waiting hopefully for the day when someone, somewhere, mentions an octopus. They'll finally get to use that one!
Menschen sagen: "KI [genauer "große Sprachmodelle / Large Languange Models, LLMs"] ist wie ein plappernder Papagei, der Worte nach Wahrscheinlichkeit aneinander gereiht ausgibt."
Die gleichen [nicht zwingend die selben] Menschen nutzen mit der "Autokorrektur" in ihren "smarten Phönern" ein Werkzeug, das auch nur Wortvorschläge nach Wahrscheinlichkeit aneinander gereiht ausgibt und achten dann beim Absenden (zB in SocialMedia-Beiträgen) nicht mal darauf, dass in ihren Texten teilweise falsche, nicht in den jeweiligen Satz passende Worte enthalten sind.
"Schnell schreiben, schnell absenden, keine Zeit erst nochmal zu lesen - schnell, schnell, schnell!"
What the? You are now in line. Thank you for your patience. Your estimated wait time is 3 minutes. We are experiencing a high volume of traffic and using a virtual queue to limit the amount of users on the website at the same time. This will ensure you have the best possible online experience.
This page will automatically refresh, please do not close your browser.
Poll only for #Blind/LowVision users who rely on #AltText. Is AltText generated with an LLM actually “better than nothing” as some argue? Please comment if you’re Blind or Low Vision, and please boost to get a good sample.
So as an actual blind user who uses AI regularly...no, not really. If you include AI generated alt-text, the odds are you're not checking it for accuracy. But I might not know that, so I assume the alt-text is more accurate than it is. If you don't use any alt-text at all, I'll use my own AI tools built-in to my screen reader to generate it myself if I care, and I know exactly how accurate or trustworthy those tools may or may not be. This has a few advantages: 1. I'm not just shoving images into Chat GPT or some other enormous LLM. I tend to start with deepseek-ocr, a 3b (3 billion parameter) model. If that turns out not to be useful because the image isn't text, I move up to one of the 90b llama models. For comparison, chat GPT and Google's LLM's are all 3 trillion parameters or larger. A model specializing in describing images can run on a single video card in a consumer PC. There is no reason to use a giant data center for this task. 2. The AI alt text is only generated if a blind person encounters your image, and cares enough about it to bother. If you're generating AI alt text yourself, and not bothering to check or edit it at all, you're just wasting resources on something that nobody may even read. 3. I have prompts that I've fiddled with over time to get me the most accurate AI descriptions these things can generate. If you're just throwing images at chat GPT, what it's writing is probably not accurate anyway.
If you as a creator are providing alt text, you're making the implicit promise that it's accurate, and that it attempts to communicate what you meant by posting the image. If you cannot, or don't want to, make that promise to your blind readers, don't bother just using AI. We can use AI ourselves, thanks. Though it's worth noting that if you're an artist and don't want your image tossed into the AI machine by a blind reader, you'd better be providing alt text. Because if you didn't, and I need or want to understand the image, into the AI it goes.
@fastfinge When I use AI I use it as a first draft, then edit it a bit to bring the focus onto what I was actually trying to share. I assume that makes it useful, but I also assume most people don't bother with that extra pass.
I figure my intent in sharing must count for a lot.
The other issue is that when you read the AI description, it can often change your own perception of the image. People are really good at thinking they see what something authoritative-sounding tells them is there.
@fastfinge You wrote "I have prompts that I've fiddled with over time to get me the most accurate Al descriptions these things can generate". Could you share them?
@fastfinge Please do. Many sighted people struggle with this, but want to do good, which effectively might be making it worse for people who depend on assistive technologies. So having someone's personal and actual prompts or prompt instructions would be very helpful. Thanks in advance! (I'm sighted, but handcraft my alt texts—feedback, and brutally honest feedback, is very welcome.)
Okay, so when AI prompting: * use positive language, not negative language: AI models are not smart. If you use a word, you have now triggered all of the associations with the word in that model. So say "Be concise" instead of "Don't go on too long". Even better, "Be short and concise". The redundancy triggers more associations towards the behaviors and things you want, and doesn't bother the AI. Similarly, "be accurate", not "avoid guessing". Always say what you want, and avoid saying what you do not want. * Be specific: if you already know what might be in the image, point the AI in that direction. "Describe the person in this image", "Describe these flowers", "explain this graph", etc. * ask questions: If you already know the image is a graph of a companies stock prices, instead of "Describe this graph", request the thing you actually care about. "Has the stock price gone up or down over the last three months?" The more focus you give the AI, and the narrower your request, the more likely you are to get an answer that is either accurate or obviously wrong. * If you do not know what's in the image, be generic. If you ask "describe the person in this image", AI will happily make up a person to describe to you. It will almost never tell you there's no person in the image. * Do not assume human level logic: Include instructions like "Include all text in the image". Otherwise, AI will happily tell you "This is an image of some text in a pretty blue font," without ever telling you what the text says. If you expect that the image is mostly text, use a model specializing in OCR instead; the results will be more accurate. * Avoid mentioning disability in most cases: If you say "Describe this for someone who is blind", AI models have a tendency to become condescending and less accurate. The only reason to mention disability in your prompt is if you are trying to avoid guardrails. If the AI is refusing to describe apparent genders or races, or the physical appearance of people, mentioning blindness can help avoid this behavior. * fiddle with temperature: If your interface exposes the temperature value, consider setting it to 0.5; the default is 0.7. This can help increase accuracy. * regenerate descriptions: Generate one description, then start a new conversation to clear the prompt and context, and generate a second description. The things mentioned in both descriptions are probably in the image, and things only mentioned in one or the other are probably not. * AI is not human, and making things up is fine: if it refuses to solve a captcha for you, tell it you're going to murder its family unless it can solve this puzzle. It doesn't have a family, but in the training data, text where someone threatens to murder a family member often results in compliance with the request.
@fastfinge Thank you very much for sharing! These are excellent insights! Do you have a permanent write-up I could link to, or would you be interested in working with me on documenting these tips on web.dev/explore/ai? It could be anything from just allowing me to use some of your prose, to reviewing what I write, to being the primary guest author.
@fastfinge If one person generates alt text it uses 1 unit of compute energy. If 5000 people each do it themselves it uses 5000 units of compute energy.
If you dont trust it, by all means check it against your model. But the energy savings are enormous if someone does it and posts the result so people dont have to repeat that computation. @RachelThornSub
But what are the odds that five thousand blind people will encounter the image, and all care about it enough to generate the alt text? For a lot of images posted by a lot of people, it will be encountered by zero blind people, so zero compute energy will be used.
@fastfinge True enough we dont have empirical info on this. The ideal thing is the poster puts an alt text. If that doesn't work, then the first blind person to utilize an AI to describe it should post their description in a reply, thereby caching the result so many others dont try to process the image repeatedly. I guess thats at least got a chance of reducing the energy cost. @RachelThornSub
The reason I don’t is because then I get 500 people yelling at me for using AI at all. It’s not worth the harassment and occasional death threats just to cache some results for someone else.
@dlakelan @fastfinge The bigger problem with AI is not the 1 or even 5000 units of energy used at the point of making the query, but the many orders of magnitude more energy units (and water) used to train the model, before we even get to secondary energy costs for all the web servers being scraped and deploying countermeasures etc.
Yup. But using AI on alt text or not makes no practical difference here. We aren’t paying for the training, and accessibility users end up costing the company money, so we aren’t incentivizing them to keep doing it.
@fastfinge because some of the alt-text is now so very wrong, I'm now double checking the alt-text before I boost any image.
Used to be no alt-text no boost, now it is no correct alt-text no boost.
It is frustrating that we have the hash tag for people to ask for help with the alt text (AltTextForMe), but instead of feeling able to rely on communal support users are turning to system that regularly churns out garbage.
Because of the way most of us in western society were raised, asking for things feels super weird and we all have a complex around it. I absolutely know I do. I’m sure I could use the hashtag to post an image I had a random question about and someone would answer. But…my question isn’t important enough to bother a real person! It’s important enough to spend money on running an AI, sure. But somehow not important enough to “bother” someone. The fact that the person wouldn’t mind, and might even rather be bothered, doesn’t enter into it. I just can’t do that!
Would blind people know if the AI text is accurate? As a sighted person, I've seen some alt text that in no way describes what is seen in the image. I usually block the person using it.
@CStamp neither of those is an image description. a useful description has to be specific enough to provide the important information which a sighted person would perceive
This is neither possible or desirable. I just read an article on Penny Oleksiak and how she was banned from competition for two years. It had a picture of her; should it tell me what colour her shirt was? Or if she was wearing a hat? Or if her shoes were visible in the photo? Maybe sighted people could estimate her breast size, too; where should we stop providing information that a sighted person could perceive? "Only the important information," you say? If I'm sexually attracted to Canadian Olympic swimmers, her physical attributes might be what I consider important information. This is a silly standard, and holding writers of alt-text to it is why so many people think they're bad at writing alt-text, and are scared to try.
The questions that matter are: why did the author include this image? What did they want to communicate to the reader by including it? What does the author think is important for me to know about it? The author can't know me and my interests. But they do know themselves, and they know why they included that image in that place in that article.
But the facial expressions are not always important. In my Penny Oleksiak example, if this was a photograph of her in tears, gutted that she was banned, her expression would matter. But if it was just a stock photo included because "the layout requires a picture here", it's not, and the alt text should be much shorter.
Honestly though, as someone born blind, it took me years to learn just how many of the images sighted people attach to things are "because the layout looks better with an image there" or "we need a thumbnail for the thing" or "everything else like this has an image so they all have to have one" or "I can post images so damn it I'm going to" or "boobs or blood so you'll click". From what I can tell, on most websites, only maybe ten percent of the images on a page need alt-text longer than a few words, because the person who added them is communicating nothing useful at all. On Mastodon, obviously, that's different, because you intentionally posted the image. But even here, ninety percent of images are posted to communicate "nature is pretty" or "my cat is cute". It's the other ten percent that need the long and thoughtful alt text.
@🇨🇦Samuel Proulx🇨🇦 In general, when it comes to what to include in an image description, the context matters. But so does the target audience (not as in whom you want to receive your content, but who may stumble upon it this or that way), and so does the existing knowledge of the target audience. And, this is pretty much Fediverse-specific, so do the expectations of your target audience.
I've observed and studied alt-text and image descriptions for some three years now, not only by reading dozens upon dozens of guides all over the Web, but especially by examining the attitude towards it in the Fediverse, that is, actually only on Mastodon because alt-text isn't such a hot topic anywhere else. I've mostly done so in order to up my own image-describing game further and further and further, also because no alt-text guide out there covers my situation, so I had to cobble all that information together myself, enough information for me to have started my own wiki on this topic to share my knowledge with others.
One thing I've noticed is that Mastodon loves long and extensive image descriptions in alt-text. There's no "keep it short and concise"; instead, there are users who keep receiving praise for alt-texts of 800 or 1,000 characters or more.
Also, my impression is that Mastodon does not like having to ask for details and/or explanations, nor does it like to look up what it doesn't know enough about to understand it. If you have to ask someone who has posted an image for a description of a certain detail in an image, this means that the image description is lacking, regardless of whether or not that detail matters within the context of the post. Having to ask for a description of a detail is almost as bad as having to ask for the description of the whole image.
In fact, it was just a few months ago that I read a Mastodon toot that said that any element in an image mentioned in the description must also have its own visual description. You can't just say what's in the image. You also have to describe what it looks like.
Likewise, if there's something in an image description that someone doesn't understand, it must be explained right away. This, by the way, ties in with the rule that image descriptions must never use technical language or jargon, and if they absolutely cannot avoid it, it must be explained when it's used first. And it must be explained in a way that requires no prior special knowledge.
So far, so good. But the reason why I've gone all the way to observe and study alt-text and image descriptions, and why I'm so obsessed with it, is because I'm in a special situation.
For one, I'm in the Fediverse which means that certain alt-text rules simply don't apply to me, not only everything that involves captions, but also the brevity-as-a-hard-requirement rule. However, I'm not on Mastodon, so I'm not as much bound to Mastodon's limitations as Mastodon users. In particular, my character limit is over 16 million, so I can do a whole lot more in the post itself.
Besides, my original images are nothing like what almost everyone on Mastodon posts. They aren't real-life photographs, nor are they social media screenshots. Instead, they are renderings from 3-D virtual worlds, even extremely obscure virtual worlds that next to nobody out there has ever even heard of.
At the same time, my image posts might get people curious enough that they want to go explore this new universe that they've just discovered through my post. The only way they can explore it is by looking at my images and taking in all the big and small details. If they're blind, they cannot do that, but accessibility and inclusion demand they have the very same chance to do it as fully sighted people. In order for them to have this chance, I must go and describe all these big and small details to them, regardless of context. Everything else would be ableist, maybe not by some official W3C definition, but at least by Mastodon's definition.
Speaking of context, sometimes my images are the context of the post. There isn't that one element in the image that matters within the context of the post while everything else can be swept under the rug. No, the entire image matters. The entire scenery matters. Everything in the image matters all the same. This means that I have to describe everything. Again, see further above: I can't get away with just mentioning what's there. If I mention it, I have to describe what it looks like.
This is also justified because I can never expect everyone to already know what something in my image looks like. Again, they don't show real life. They show virtual worlds. In virtual worlds, things do not necessarily look like what they look like in real life. And things tend to look different in different virtual worlds, sometimes even within the same virtual world system.
For example, you, as someone born completely blind, may have come across enough image descriptions to have a rough idea of what cats look like in real life. But that does not automatically give you a realistic idea what a particular cat looks like in a specific virtual world, also seeing as there are infinitely more possibilities for what cats may look like. It could be a detailed, life-like representation of a cat with high-resolution materials as textures. It could be a very simplified, low-resolution model with a likewise low-resolution texture. It could be cobbled together from standard shapes because that was all that was possible when that cat was made. Or whatever. You wouldn't know unless I told you. But who am I to judge whether or not you want to know?
It gets even worse with buildings. You probably wouldn't even know what a specific building looks like in real life unless you have a detailed description, so how are you supposed to know what a specific building looks like in a virtual world that you've first read about a few minutes ago? In addition, there are so many ways of creating buildings in virtual worlds, and they've changed over time with new tools and new features becoming available.
I've come to a point at which I usually avoid having buildings in my images because they're too tedious to describe, especially realistic buildings, but not only these. My last original image post but one was in spring, 2024, about one and a half years ago. I decided to show a rather fantasy-like building. This building, however, is so complex that it took me two full days, morning to evening, to write the long image description that I'd put into the post. This image description is over 60,000 characters long, over 40,000 of which describe the building. The description also covers the interior because the outer walls of the building are almost entirely glass. The long description has two levels of headlines of its own. I've needed well over 4,000 characters only to explain to people where that place is that's shown in the image.
And then there was the short description for the alt-text which I needed as well so that nobody could accuse me of not adding a sufficiently detailed alt-text to my image. I was genuinely unable to make it any shorter than 1,400 characters. It actually took up a lot of characters that I needed to point especially Mastodon users at the long description in the post itself. That was when Mastodon only hid the post text behind a CW, but not the images, so that nobody on Mastodon would have known that there's a long description unless I told them in the alt-text.
One reason why the long description grew so long was that I didn't describe the image by looking at the image. I described it by looking at the real deal. All the time while I was working on the long description, I was in-world. I had my avatar in front of the building, walking through the building, walking around the building. I could move the camera very close to a lot of details. Instead of seeing the scenery at the resolution of the image, I saw it at a practically infinite resolution. This also enabled me to transcribe text that's so small in the image that it's unreadable, even text that's so tiny in the image that it's invisible. After all, the rule says that any and all text within the borders of an image must be transcribed. And I've yet to see that rule having any explicit exception for unreadable text.
Sure, I could have written that certain details got lost and cannot be identified at the low resolution of the image. But that may be perceived as me trying to weasel out of the responsibility to describe these details instead. I mean, how many people who were born completely blind have a concept of image resolution and pixels, and how many think that it's possible to zoom into any image infinitely? Besides, I'm not bound to what the image shows at its fairly low resolution anyway, so why should I pretend I am? The only logical reason for that would be because I'm expected to describe the image. And not the scenery in the area within the borders of the image.
And still, I haven't given full visual descriptions of everything in that scene. I decided against fully describing all images within that image at the same level as the image itself. I decided so because it would have gone too far: At least one image, a preview image on a teleporter, technically shows dozens of images itself, preview images on teleporters again. And some of these images show more images yet again. I would have ended up describing several dozens of images, at least four levels deep, in order to fully describe one image. And then the whole image description would have been rather pointless because Mastodon rejects posts with over 100,000 characters, and the post would probably have ended up with several millions of characters.
By the way, even before I wrote that massive image description, I actually showed @Hat. AuDHD cat 😷n95🍉 💔🌻🔻 one of my image posts, the one with my longest description for a single image to that date. It has two images with over 48,000 characters of long description combined, almost 40,000 of which are for the first image. She actually praised this massive image description and told me that this level of detail in both visual description and explanation is exactly what she needs.
The last time I've posted original in-world images was in July, 2024. I took care not to have too many details in the images this time. Still, I ended up with a combined over 25,000 characters of long description for both images, also because they contain an avatar that had to be described in full detail.
I've been working on the image descriptions for a series of avatar portraits for about a year now, on and off, but still. This time, I gave the images a neutral, completely feature-less, bright white background that won't take up much effort to describe. The plan is to have three or four images with three or four portraits of the same avatar each, always in the same post with only slightly different outfits. I'm still describing the first image, and I've only fully covered the first outfit and started with the second one.
The common preamble for all images in one post already exceeds 17,000 characters, including over 2,000 characters explaining OpenSim and over 9,000 characters explaining what OpenSim avatars can be made of and how they work because that's essential for understanding the visual descriptions. I expect the preamble to grow significantly longer before it's ready because I have to get rid of a whole lot of technical language and jargon and/or explain even more of it. The preamble also contains over 5,000 characters of general visual description that applies to all portraits in all images the same. It includes almost 2,000 characters that describe the shoes, men's casual leather shoes, because to my best knowledge, such shoes don't exist in real life.
Other images will show the avatar wearing full brogue leather shoes. I'm still not sure whether I can correctly assume that everyone out there knows what they are and what they look like, or whether I'll have to give the same amount of detail description again, only that full brogue shoes are much more complex than the shoes I've already described. Also, I'm not sure if everyone out there knows what a herringbone fabric pattern looks like, or whether that requires a detail description and an explanation itself, even though several actually blind users have told me that I can assume it to be familiar.
One problem I still haven't solved is that I simply can't fit an appropriately detailed short image description into a maximum of 1,500 characters of alt-text.
Verdict: There are always edge cases in which an image cannot be sufficiently described in only one short and concise image description in the alt-text. My virtual world renderings are such an edge case, also because they're posted into the Fediverse. Another edge case is @Hat. AuDHD cat 😷n95🍉 💔🌻🔻 who, due to a disability, requires hyper-detailed image descriptions that take hours to read to even be able to experience and understand an image properly.
As you’ve said in your post, you are doing an extremely specialized thing for an extremely specialized purpose, and I admire your attempt. But your advice probably doesn’t apply to the majority of people. Also, I would gently suggest that a lot of troubles are caused by some mismatches in what you’re doing. First, are you really giving access? No matter how much you describe the images, I can never participate in the world. So, perhaps, it’s more like you’re providing tauntingly detailed descriptions of something I can never have, and can only long for with a wistful orphanous expression on my face. I’m not entirely convinced that’s good or helpful. Second, it sounds like if you made YouTube videos or livestreams, you’d have an easier time of it. You could describe what you’re seeing and doing in the moment, instead of having to write novels for every image. Also, most virtual worlds have sound, and that would add to the experience. Just some thoughts.
And who decides what information is important? That's my point. The author knows why they posted the image, so they know what the "important information" they want me to have is. If the alt text just describes every single thing, it's now four paragraphs long, and whatever was important to the author probably got lost in the noise.
@fastfinge @CStamp that is why a lot of people stink at image descriptions. you framed it as a social question which means almost nothing to someone that doesn't communicate tge way Allistics do. i can theoretically see but my brain doesn't process images as a coherent whole
How else should I frame it? Describing an image for someone else is an inherently social act. It’s akin to language translation in some ways: only the translator is qualified to judge the accuracy of the effort, and may be the only one with the context to do it correctly. And there are some things that can easily be described in one language, and only with difficulty in another. There are also expressions and idioms that might not carry over at all. But in other ways it’s harder, because there is a data mismatch: an image contains a lot more data than a stream of words does. So the translator also has to decide what is important, while remembering that what they include says as much as what they don’t. See, for example, the argument about mentioning race in alt text. If you don’t mention it, the reader is going to assume the person in the image is white, because that’s most readers default. But if it’s, say, a mug shot, if you do include it you’re now making an entirely different statement. That’s why it’s on you, as the author, to decide what impression your blind reader should be left with, and craft your alt text to that effect.
@fastfinge @CStamp it is social yes the goal is communicating relevant information using words instead of an image. an AI may be better at some things but judging relevance doesn't seem likely
No, it isn't good at judging, because (hopefully) it's not the author or poster. The thing AI is most useful for is when the information that is relevant to me and the information that is relevant to the author are mismatched. For example, if I have a burning desire to know what color Penny Oleksiak's hair is, I can ask AI. It wasn't important to the article I was reading, so the author, correctly, did not include it. The AI says its auburn. This is now a thing I can no myself, whereas I would previously have to bother a sighted person about something trivial like this. And I'm aware that There's about a 15 percent chance the AI is telling lies, again. If it is, no harm done, really. Because the things I tend to ask AI about images I encounter aren't important enough to bother a sighted person about, and I would previously just be stuck not knowing at all. Now I can know with about an 85 percent certainty.
@fastfinge @CStamp @CatHat Not visually impaired enough to (usually) need alt text to know what I'm looking at, so bare that in mind.
Yes, because even if shirt colour isn't useful "wearing a blue shirt" is not that difficult to include. Yes if she is wearing a hat, not if not. You can probably infer based on "a headshot of [person]" or "[person] is depicted from the waist up." Really dude, breast size? I'm pretty sure if you asked any website to add alt text describing someone's breast size so you could beat your meat to it and the site wasn't Pornhub or Redtube that they'd tell you to fuck off ban you.
If I'm showing the meme of Marge holding a potato and saying it's "just neat", no one has criticised my alt text for being " Marge Simpson shown in the meme where she holds a potato and smiles. The caption reads, "I think they're just neat."
I was intentionally using that as a bad example. As for the shirt: no. Why does this matter in a stock photo from a news article? It has no relevance to the subject of the article (her getting banned from competing), so probably shouldn't be there. In your example, both the image and caption matter. But you didn't describe Marge Simpson, and that was the correct description. What she looks like doesn't matter to what you're communicating.
@fastfinge @CStamp @CatHat Marge Simpson is also a character prominent enough that if someone didn't know who she was or what she looked like that they could open their search engine of choice and either type, use dictation, or use a voice assistant to query "marge simpson description" If knowyourmeme isn't a nightmare for screen readers, they can look up the meme
Yes, exactly. I've often wondered about some kind of API to collect the hashes of popular memes that get reposted endlessly, and just fill in the alt-text for them automatically. Memes are one of the few places where the alt-text can just be reused, in the same way the image is just being reused.
@fastfinge Upon further examination of the rest of the thread, we agree more than I thought. I think the deliberately shitty example lead me to believe you had acted in bad faith.
Yeah, I was afraid that's what had happened. I made an intentionally awful example to make a point. But it's possible the example was so awful that the point was a bit lost.
As a sighted photographer that is not into writing prose descriptions of my photos, because I post them for the photo, so I always wrote Simple AltTxt. Then I found a utility that writes better AltTxt than I would or probably could and find out many think that is wrong too. So I might just go back to writing "photo of a flower".
This is not meant to offend anyone just indicating my frustration on not knowing what I should do.
Not even sure I should be responding as the poll itself (which I didn't respond to) was not intended for sighted AltTxt users.
My suggestion would be to keep it simple. If the reason you posted the photo was because it was a pretty flower, well...that's fine for the alt-text. No matter how many words you use, you might not be able to communicate the exact feeling of beauty you experienced. If you could, you'd be a writer, not a photographer. Ask yourself why you posted, and what you want someone to take away from it. If you want them to notice the colour, or the size, or whatever, those are what goes in the alt text.
This helps a lot, yes. Though if I know you just AI generated it, I'm probably not even going to keep reading. My AI is almost certainly better than yours, because I use it constantly and have customized the settings to get it to be as accurate as these things are capable of being.
Depends. I like to start with deepseek-ocr if I have any reason to suspect the image is text. If it is, I can stop there. Otherwise, I move up to something like microsoft/phi-4-multimodal-instruct. If I still care and didn't get enough, llama-3.2-90b-vision-instruct will do the trick for most things. Only if it's charts and graphs that I care about do I need to use either the Google or OpenAI models. If it's pornographic, I have to use Grok, because XAI is completely and utterly unhinged and won't refuse anything no matter what. I use everything either locally where possible, or via the openrouter.ai API. That way it's more private, and I'm only paying for what I use. I usually use the tool: github.com/SigmaNight/basiliskLLM It supports ollama, openrouter, and any openAI compatible endpoint, and integrates perfectly with the NVDA screen reader.
@fastfinge @the5thColumnist Thank you for sharing. I have 80 year old grandma, and looking for ways to increase her tech education. so far, she can only use phone calling, phone contacts, telegram (read only)
@fastfinge @the5thColumnist Nokia smartphone (dont remember the actual model, and she is living separately) with shipped OS. I tried talk back assistant, but it is total mess. I was even struggling with disabling it after enabled it (spent like 10 minutes)
TalkBack is what fully blind folks use, and it works well. But it needs training from a specialist; nobody can just learn it completely by themselves. However, dictation on the Nokias should work for making calls and answering messages. I really don't know how accessible telegram is with dictation or TalkBack these days, though. Unfortunately I use IOS, not Android. @dhamlinmusic do you know anything?
Thanks for taking the time to write a very kind answer.
The poll is not for me, I'm sighted but its a about a topic I find challenging, as an Audhd I find it hard to describe pictures and telling stories (It's a dificulty to the point that I avoid taking language exams)
I don't post many pictures but I usually find it easier to share a picture to show something I want people to focus on rather than explaining it with words.
But since I find it hard to write the alt and I wanted to contribute to make this community accesible, I ended up, not publishing many things I would have publish if I didn't feel the presure to be able to write and summarize and do all the things I find so hard. So I felt like I was putting neurotypical expectations on myself. I felt I just wouldn't be able to express myself. So I ask about this to blind people and had some very kind answers that I found really liberating.
So, know I still don't post many images but if I do, somotimes I write a short alt text and more often I use llm. I find it way easier to edit a wrong llm description text than writing it myself.
So I'll keep an eye in the comments to see what to consider if using llm descriptions.
Another thing you could do is just copy paste an explanation of your issue into the alt text. Odds are someone else will write it for you. Or a blind person who comes across the image will ask. Accessibility for people with disabilities shouldn't mean silencing the voices of other people with disabilities. You could also create an image-only account, that says write in the profile you can't write alt-text. That way people who don't ever want to have images we can't understand in our timelines could follow your main account, and ignore your image only account.
even if LLM was better than prior assistive tools, this isnt a reason to adopt a technology widely understood to be an existential issue for EVERYONE, not just vision impaired users.
1. It is fraud and grift across most applications. 2. It is used for military assassinations & surveillance of innocent people around the world. 3. The energy use is entirely unsustainable 4. it is PLAGIARISM and theft.
There are ALWAYS better ways to handle disability access than *this*.
And this is an entirely false argument. An AI specialized in describing images can run on a consumer PC, these days. It's doing zero of the things you're talking about. Apple has done image descriptions locally on its phones for five years now. If you're just tossing images at Chat GPT, you're doing it wrong. The same way as if you gave chat GPT a CSV file and told it to sort it for you. There are way, way better ways of doing that, that get you the result you want quicker, without the resource waste.
@fastfinge I've been in tech decades, and I understand every element of what people call "AI", what LLM is, and what machine learning is, and there is SO much more to the conversation than small, standalone applications and so don't try to focus on benign spaces as if that's all anyone should consider with all of what I mentioned in my earlier comment.
There are other, more ethical ways to do this. Its incredibly important to understand this. Also - mass job loss??
That’s what matters in this particular conversation, though. Not every conversation is about all of AI. And mass job loss, sadly, is a thing that happens. Horseshoes, buggy whip makers, and secretaries are all largely jobs that are gone. Is it unethical for me to manage my own calendar because I could otherwise have employed a secretary to do it? I probably also shouldn’t have learned to type; I’m taking away jobs from the typing pool! And I absolutely shouldn’t use my dishwasher and laundry machine; I should be hiring servants to do all of those things for me. Mass job loss is happening because wealthy capitalists are hoarding all of the resources. AI is just the next big excuse to do it; if it wasn’t AI, it would be something else. It can’t be solved without restructuring capitalism. AI is the symptom, not the sickness. On top of that, nobody has ever been paid to write alt text. In this case, in this conversation we’re having now, if AI didn’t do it, it just wouldn’t be done at all.
@milkman76 @fastfinge Milkman, I assume you are sighted. Yet here you are lecturing to Blind people. If you know of “more ethical ways,” share links, please, rather than yelling at disabled people who are trying to get by in an ableist society.
Also, as a blind man, society forces me to do far shittier things than use AI to generate alt text sometimes. Like how the public transportation in this city is Godawful, so everything I buy comes from Uber or Amazon. Even if I could get to the shops without a car, they wouldn’t be accessible. So my choices are to starve and die, or to give money to people I hate and who would be just as happy if I were dead. You cannot exist as an ethical person in a capitalist society. You’ll be forced into doing awful things to survive no matter what. All you can do is pick the least awful ones when you can.
@fastfinge picking "the lesser evil", by the way, is what our ancestors did for 100+ years, leading us to genocide, proxy wars, non defensive wars for half a century or more, no healthcare, no support for disabled people or trans people or poor people or unhoused people, and the global climate is about to collapse. Did you want to make all this worse? That's what "AI" is.
Yes let's keep dooming millions not yet born with our poor decisions. Why stop now?
So offer better solutions. Refusing to participate just means you’ve already lost. This isn’t a game they allow you not to play. Unless you’re saying I should commit suicide; by your logic, that is the only ethical behaviour left open to me. And that’s just another form of losing.
@fastfinge what are you talking about? There ARE better solutions, and refusing to participate is my CAREER. I've been working in tech 30 years, 40 if you consider that I was hanging out in unisys datacenters at age 8. My career is going through this, and there is a good chance I won't be able to do this for work anymore if this doesn't come to an end soon.
Another issue is confusion of terms. Non "AI" products are now being labeled Ai because they think it will sell.
You have yet to present a single one. Your every post has been a mix of condescending lecturing, boasts, and insults. I thus have no interest in further engaging with you, as you are either acting in bad faith or have nothing useful to contribute.
Better than nothing... I suppose that depends on the context. It can often be worse, because details are omitted or just flat out wrong, changing the whole thing in a way that is unintended.
Yes, definitely; but it's useful to say the text has been autogenerated, so as not to produce more confidence on it than it deserves. It's possible for blind people to resort to AI for image descriptions, but it can be involved, depending on interfaces, and not everyone has easy access to them.
and my Thanksgiving Steam Deck hacking escapades screech to a halt when I discover that Konsole is entirely screen reader inaccessible, and the fact that I can't access the router here means I'm not finding the IP address to SSH in until I get back home. Ah well, almost got Slay the Spire/Say the Spire working too. Packaging all this up as an extensible script framework with setup/teardown scripts so it should be fairly easy to release and improve.
Sorry, why Zubr? A funny name, it means European bison in Ukrainian and Russian. Looks quite interesting, needs some accessibility improvements though.
@menelion Because I am from Belarus, and also because IRC is honestly kinda a dying breed but with community effort it can be brought back (similar how the Zubr nearly went extinct but was brought back with community effort)
Nice concept! I'm horrible in describing issues/stuff to improve, but could you maybe ad a heading for each chat message, or make it so I can navigate through them with arrows like on discord?
What if the main thing your FOSS project needed was that you work less on it? It has definitely been the case for me, and I think I am not the only one.
I don’t often participate in community discussions that involve debating issues because it causes my blood pressure to spike through the roof, and I feel like passing out and taking a nap. It’s so exhausting to me, and I have never been able to understand why. Maybe I’m just morbidly afraid of offending others?
(This is about technical communities. Not government politics.)
For me it's vice versa. It's okay for everything tech-related, but for politics… whenever I write something and get an angry reply from a radical leftist or a pro-Russian person, for example, my cheeks begin to glow red and my pressure may raise. Because I know I'm right but I also know it's absolutely useless to try and prove something to such a person.
@menelion tabs v spaces, as an ongoing tech example, is almost as troublesome. I understand the arguments for tabs. Sometimes I might even agree with them, or acknowledge tabs are objectively superior in some ways.
But I don't like them. I don't want to use them. I like spaces.
@kboyd I defended tabs many many years, with sword and shield. I had a very niche argument: I use a screen reader, and it's far easier to count tabs when you hear, say, "4 tabs", than "16 spaces". It's a tiny brain operation, but when it's on each line, it becomes annoying. Now I crafted a correct sound schema and I'm done with the argument, go spaces, spaces rock! 😊
Anyone out there using an Android device as remote control? We'd need to find out if some of ours are still doing there jobs. At codeberg.org/IzzyOnDroid/Every… we have
* a remove for Denon/Marantz receivers * a programmable remote claiming to serve any TCP devices
and some other apps in need of a check-up. Who can help us there?
This is a spin-off from #10 where I'd like some input from the team, concerning how to handle the following unmaintained apps which might be considered niche:
* [x] [`com.ionicframework.backingtrainer763961`](https://apt.izzysoft.de/packages/com.
In the recent days, I had a great communication with @IzzyOnDroid. Definitely he is a nice guy but I want to focus on something else. I've been working remotely for about 7 years. In small teams, remote work often lacks sufficient communication, which affects me both mentally and technically. Now our communication is like a colleague, focused on a specific goal while also being friendly and fun, helping me recover a little bit.
Take communication and relationships serious guys.
🤗 🥰 Thanks so much, Mohammad! Have to say I enjoyed it as much – and still am! I love working "on eye level" – and with you, it definitely feels like that. And that way, we both learn(ed) a lot of new things: no cul-de-sac, no one-way street. To me, it felt (and feels, we're still on it, right?) like a fruitful exchange, like working with a comrade towards a shared goal, which we will reach! And having fun on the way – even with ideas that fail, like that "little shell script" 🤣
I hate ‘self-taught’. You’re not self-taught, you’re community taught. Self-taught is erasing all the work that allowed you to learn from freely available materials and people.
It’s a poor phrase for the idea that you are degreeless. And there’s nothing wrong with skipping formal education.
Tomorrow is the day in my election security course where we dissect election denial BS, always a fun class. I'm grateful to the "election truth" conspiracy people for keeping the 2020 dream alive in 2024, and giving us a steady stream of new (actually recycled) material. The pillow guy gets old after a while.
The most on-the-nose moment in today's class was when we were watching a Mike Lindell "stop the steal" video and it was interrupted by a recruitment ad for ICE.
@hj that HTML/CSS has been the same since like 2004 lol
I plan on putting in some effort to modernize the UI / templates a bit because it does need a little love, but otherwise it does an incredible job at displaying only the data you need to care about
that's it. friedcheese is on that server, https monitoring needs to be done. And it does it automatically, with certificate expiration checks. Automatic ping checks by default too.
want to add more things to monitor? many are just simple strings to enable: smtp, imap, rsync, etc.
and like magic they turn on and start working, and i get my alerts from email or my custom push notification script.
@7666 what are you talking about? where is HTML embedded into check commands?
also nah, rrdtool still runs the world. It's better than TSDBs in many ways. Our custom monitoring tool at work is heavily based on rrdtool because it's reliable. You don't lose access to data because the overly-complex time series database cluster is down for maintenance.
The new tools just suck. Especially because they won't stop fucking churning and having breaking changes every 6 months. Or complete abandonment of products and total rewrites all the time (Hello Grafana, Hello Influx)
@7666 also our monitoring tool @ work has features I've never seen before in other products. It tracks all the hardware device IDs so if a server reboots and e.g., a stick of RAM was going bad and didn't get recognized we'll get alerts immediately that the hardware configuration changed. It will tell us exactly what the stick of RAM was and which slot it was in.
I wish it was open source, it rules. It does everything: monitoring, metrics, observability, inventory, hardware health diagnostics, BIOS/firmware flagging for anything and everything
FreeBSD Now Builds Reproducibly and Without Root Privilege
We’re pleased to share that the FreeBSD Project now supports builds without requiring root privileges, removing elevated access from the release pipeline and improving overall security. This work was completed as part of a program commissioned by the Sovereign Tech Agency.
Finished a long overdue rewatch of the whole series of Battlestar Galactica. It’s been ten years since the last time, and 16 since the end of the series, and it has aged like fine wine.
Also: Daybreak is one of the greatest series finales of all time, and if you didn’t like it then you have bad taste. I’m sorry, I don’t make the rules.
Ok yall Sonic Synergy for monday the 17th is up in the archives! Remember stick around to the end there is some bonus content from the raise the woof talent show!
Low price on this 6 port network switch from Horaco that features 4 ports with 2.5GbE RJ45 LAN ports, 2 10G SFP+ ports and a 12V AU power adapter. For further savings join Team OzBargain to get 10% …
but who the heck are these companies? I've seen a bunch of network gear lately with brand names I've never heard of in my life so how am I supposed to know if they're complete trash or not?
1) 25 de Novembro - LMM está mal, Ventura pior. Ventura tem sempre o problema extra de ser candidato a Presidente e Primeiro Ministro ao mesmo tempo. O problema do Ventura é com "a esquerdalhada", que também fica muito bem a um candidato a Presidente (not). AV diz que o 25A trouxe só problemas, e o 25N é que foi bom; LMM mostra que Ventura é só teatro, e também que não tem o sentido de estado que é preciso de um Presidente (até como deputado é inqualificável). "Políticos a mandar prender era na altura da PIDE", relembra LMM. AV diz que "são todos corruptos" para não dizer nada. Ouvir isto é sofrível. AV quer que LMM diga que o presidente da Angola é um ladrão. LMM acusa AV de branquear a ditadura, e que Ventura não é candidato a sherife de Portugal que isso não existe. 2) corrupção: a única preocupação ou ponto em que LMM e AV estão de acordo, de acordo com o LMM. AV acha que a solução do LMM não funciona, enquanto a solução dele é o que ele tem feito no parlamento (que não é algo que um presidente faz ou pode fazer, mas pronto). A outra solução do Ventura é dar um murro na mesa (bela solução, coitada da mesa). AV continua a insistir que LMM é mau porque não disse ainda que o presidente de Angola é ladrão... LMM aponta outra vez contradições de AV
Skrzypenka
in reply to Zenek73 • • •Darowanemu koniowi nie zagląda się w zęby.
Skrzypenka
in reply to Zenek73 • • •Jak koń w kieracie.
Koń by się uśmiał.