#imagedescription — Public Fediverse posts
Live and recent posts from across the Fediverse tagged #imagedescription, aggregated by home.social.
-
@Ruby SkunkloonFor both alt text and CWs, I wish there was a more automated way to handle this. Like, having canned CW tags at the ready, or an image detection bot that actually works as intended for filling out alt text. (That's before getting into the weeds of "AI", which is contentious in itself.)
For CWs, in theory, things could be a whole lot better. A whole lot better than what Mastodon's culture demands.
Instead of the writer supplying a CW and forcing the self-same CW upon all readers all the same, CWs could be automatically generated on the readers' side. Like, those who need that particular CW get it, and those who don't need it don't get it.
This is technically possible. Mastodon has introduced it with version 4.0 in October, 2022. It's a filter setting. In general, this has been available in the Fediverse since 2010 when Friendica introduced it first. On Friendica and all its descendants, CW culture has always been built around this rather than writing CWs into the summary field and forcing them upon everyone all the same.
This has got a few downsides, though:- Writers still have to remember to take care of adding filter-triggering keywords.
- The best way to add keywords without shoehorning them into the text is as hashtags, but many Mastodon users have a problem with more than, say, four hashtags per message.
- Whereas Friendica and its descendants simply have a keyword list for this purpose, Mastodon requires one new filter to be set up for each keyword.
- I guess there are lots of Mastodon apps that don't have creating or editing filters built in. And I know there are lots of Mastodon users who would never ever touch Mastodon's Web UI under any circumstances.
- Mastodon users are used to being coddled and pampered all over. Most of them would never take care of CWs being generated on their side by themselves. They expect CWs to be served to them by the writers.
- And then there are those Fediverse server applications that either don't have a hiding/CW-generating filter or no filters at all.
Fully automated, zero-click, fire-and-forget image description generators have been desired by Mastodon users for about as long as alt-texts have been a thing. But let me tell you: Fully automated, fire-and-forget image description generators that can describe any image perfectly accurately are utopic. Even more if they're also expected to adhere to all existing image-describing and alt-text rules and adapt the descriptions both to the context and to the audience that's to be expected.
Trust me. My original images could barely be more obscure and niche. They're renderings from extremely obscure 3-D virtual worlds. And I've actually had two of them described by an AI after writing my own highly detailed image descriptions 100% by hand with no AI assistance. It took me several hours to write one description and two days to write the other one.
The results were abysmal. The AI descriptions were no match for my hand-written descriptions at all. They omitted things that really stood out in the images, they got things wrong, they guessed things that I knew for a fact.
Granted, it was extra unfair because the AI only had the image to look at whereas I wrote the image descriptions by looking at the real deal in-world with an almost infinitely higher resolution, with the ability to look around obstacles even. Still, at least for my own images, I consider this style of writing image descriptions the benchmark. AI will never get there. It will never have the extreme niche knowledge of an expert, and it will never be able to see things that I see.
CC: @Princess Nettles Succubish
#Long #LongPost #CWLong #CWLongPost #FediMeta #FediverseMeta #CWFediMeta #CWFediverseMeta #Mastodon #MastodonCulture #CW #CWs #CWMeta #ContentWarning #ContentWarnings #ContentWarningMeta #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta #AI #LLM #LLMs #AIVsHuman #HumanVsAI -
@Ruby SkunkloonFor both alt text and CWs, I wish there was a more automated way to handle this. Like, having canned CW tags at the ready, or an image detection bot that actually works as intended for filling out alt text. (That's before getting into the weeds of "AI", which is contentious in itself.)
For CWs, in theory, things could be a whole lot better. A whole lot better than what Mastodon's culture demands.
Instead of the writer supplying a CW and forcing the self-same CW upon all readers all the same, CWs could be automatically generated on the readers' side. Like, those who need that particular CW get it, and those who don't need it don't get it.
This is technically possible. Mastodon has introduced it with version 4.0 in October, 2022. It's a filter setting. In general, this has been available in the Fediverse since 2010 when Friendica introduced it first. On Friendica and all its descendants, CW culture has always been built around this rather than writing CWs into the summary field and forcing them upon everyone all the same.
This has got a few downsides, though:- Writers still have to remember to take care of adding filter-triggering keywords.
- The best way to add keywords without shoehorning them into the text is as hashtags, but many Mastodon users have a problem with more than, say, four hashtags per message.
- Whereas Friendica and its descendants simply have a keyword list for this purpose, Mastodon requires one new filter to be set up for each keyword.
- I guess there are lots of Mastodon apps that don't have creating or editing filters built in. And I know there are lots of Mastodon users who would never ever touch Mastodon's Web UI under any circumstances.
- Mastodon users are used to being coddled and pampered all over. Most of them would never take care of CWs being generated on their side by themselves. They expect CWs to be served to them by the writers.
- And then there are those Fediverse server applications that either don't have a hiding/CW-generating filter or no filters at all.
Fully automated, zero-click, fire-and-forget image description generators have been desired by Mastodon users for about as long as alt-texts have been a thing. But let me tell you: Fully automated, fire-and-forget image description generators that can describe any image perfectly accurately are utopic. Even more if they're also expected to adhere to all existing image-describing and alt-text rules and adapt the descriptions both to the context and to the audience that's to be expected.
Trust me. My original images could barely be more obscure and niche. They're renderings from extremely obscure 3-D virtual worlds. And I've actually had two of them described by an AI after writing my own highly detailed image descriptions 100% by hand with no AI assistance. It took me several hours to write one description and two days to write the other one.
The results were abysmal. The AI descriptions were no match for my hand-written descriptions at all. They omitted things that really stood out in the images, they got things wrong, they guessed things that I knew for a fact.
Granted, it was extra unfair because the AI only had the image to look at whereas I wrote the image descriptions by looking at the real deal in-world with an almost infinitely higher resolution, with the ability to look around obstacles even. Still, at least for my own images, I consider this style of writing image descriptions the benchmark. AI will never get there. It will never have the extreme niche knowledge of an expert, and it will never be able to see things that I see.
CC: @Princess Nettles Succubish
#Long #LongPost #CWLong #CWLongPost #FediMeta #FediverseMeta #CWFediMeta #CWFediverseMeta #Mastodon #MastodonCulture #CW #CWs #CWMeta #ContentWarning #ContentWarnings #ContentWarningMeta #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta #AI #LLM #LLMs #AIVsHuman #HumanVsAI -
@Alex R @✨pencilears✨(KD7POW) @Princess Nettles Succubish There's so much that next to nobody on Mastodon knows about alt-texts.
For example, how inconvenient long alt-texts are for screen reader users. Or what a screen reader is in the first place.
Or that there are people who cannot access alt-texts, and therefore writing around character limits by putting explanations and other information exclusively into the alt-text counters accessibility. (In fact, that there are also other disabilities and disorders present in the Fediverse than visual impairments.)
Or that Misskey and its various forks and forks of forks (Sharkey, Iceshrimp-JS, Catodon, CherryPick etc.) have an alt-text character limit of only 512. That they're supposed to truncate longer alt-texts at this mark. And that, due to a bug that still hasn't been fixed, they silently delete such alt-texts instead, giving the impression that there hasn't been an alt-text to begin with. (Or that Misskey and its various forks and forks of forks exist in the first place.)
The attitude is rather that any and all information that someone may need about an image must immediately come with the post. In the alt-text, of course. Explanations included. Because, you know, having to ask for explanations or having to ask for a description of a detail is almost as bad as having to ask for an image description in the first place. Such appears to be Mastodon's alt-text culture.
Those affected the most are people like me. People who only ever post images about super-obscure niche topics. In my case, that's renderings from extremely obscure 3-D virtual worlds, memes about these virtual worlds and memes about the Fediverse (not limited to memes about Mastodon). They all require extensive explanations, and the former require extensive visual descriptions, at least judging by what Mastodon's culture seems to demand. I mean, how is anyone supposed to know what anything looks like in my images?
At least I don't really have any character limits for message texts to worry about, so my long descriptions and explanations can go into the message text of only one single post. Still, I think we have proof now that the alt-text police demands a sufficiently descriptive alt-text even when the image is extensively described in the message text.
I guess nobody on Mastodon has ever really had a problem with me filling Mastodon's alt-text character limit up to the very last character. They'd rather mute or block me for exceeding 500 characters in the message text.
At least all this rarely happens. Describing and explaining images while trying to adhere to all the rules and guidelines out there, Mastodon's culture included, has become so tedious and time-consuming that I haven't posted a single non-meme image in over two years now.
#Long #LongPost #CWLong #CWLongPost #FediMeta #FediverseMeta #CWFediMeta #CWFediverseMeta #Misskey #Forkey #Forkeys #Sharkey #Iceshrimp #IceshrimpJS #Catodon #CherryPick #CharacterLimit #CharacterLimits #CharacterLimitMeta #CWCharacterLimitMeta #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta #ScreenReader #ScreenReaders #MastodonCulture -
@Alex R @✨pencilears✨(KD7POW) @Princess Nettles Succubish There's so much that next to nobody on Mastodon knows about alt-texts.
For example, how inconvenient long alt-texts are for screen reader users. Or what a screen reader is in the first place.
Or that there are people who cannot access alt-texts, and therefore writing around character limits by putting explanations and other information exclusively into the alt-text counters accessibility. (In fact, that there are also other disabilities and disorders present in the Fediverse than visual impairments.)
Or that Misskey and its various forks and forks of forks (Sharkey, Iceshrimp-JS, Catodon, CherryPick etc.) have an alt-text character limit of only 512. That they're supposed to truncate longer alt-texts at this mark. And that, due to a bug that still hasn't been fixed, they silently delete such alt-texts instead, giving the impression that there hasn't been an alt-text to begin with. (Or that Misskey and its various forks and forks of forks exist in the first place.)
The attitude is rather that any and all information that someone may need about an image must immediately come with the post. In the alt-text, of course. Explanations included. Because, you know, having to ask for explanations or having to ask for a description of a detail is almost as bad as having to ask for an image description in the first place. Such appears to be Mastodon's alt-text culture.
Those affected the most are people like me. People who only ever post images about super-obscure niche topics. In my case, that's renderings from extremely obscure 3-D virtual worlds, memes about these virtual worlds and memes about the Fediverse (not limited to memes about Mastodon). They all require extensive explanations, and the former require extensive visual descriptions, at least judging by what Mastodon's culture seems to demand. I mean, how is anyone supposed to know what anything looks like in my images?
At least I don't really have any character limits for message texts to worry about, so my long descriptions and explanations can go into the message text of only one single post. Still, I think we have proof now that the alt-text police demands a sufficiently descriptive alt-text even when the image is extensively described in the message text.
I guess nobody on Mastodon has ever really had a problem with me filling Mastodon's alt-text character limit up to the very last character. They'd rather mute or block me for exceeding 500 characters in the message text.
At least all this rarely happens. Describing and explaining images while trying to adhere to all the rules and guidelines out there, Mastodon's culture included, has become so tedious and time-consuming that I haven't posted a single non-meme image in over two years now.
#Long #LongPost #CWLong #CWLongPost #FediMeta #FediverseMeta #CWFediMeta #CWFediverseMeta #Misskey #Forkey #Forkeys #Sharkey #Iceshrimp #IceshrimpJS #Catodon #CherryPick #CharacterLimit #CharacterLimits #CharacterLimitMeta #CWCharacterLimitMeta #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta #ScreenReader #ScreenReaders #MastodonCulture -
@✨pencilears✨(KD7POW)I think a lot of people have unrealistically high expectations for alt text, and as a consequence they add a bunch of fluff, which makes writing alt-text feel daunting if that's the standard.
Looking at this phenomenon from outside of Mastodon (I'm on Hubzilla), I've got the impression that this has become engrained into Mastodon's very culture.
Mastodon has a limit of 1,500 characters so that the text in images with a lot of text can be transcribed in alt-text. However, many Mastodon users, knowing everything they know about online accessibility, take this as an invitation to write very detailed visual descriptions. Other Mastodon users cheer and applaud for these long descriptions.
Maybe it's just my impression. Maybe I'm wrong. But I kind of expect at least some Mastodon users to start both demanding and enforcing very long and detailed alt-texts/image descriptions, regardless of long alt-texts actually being bad for various reasons and even Mastodon users saying so.
I mean, I've already read several times that images must also be explained if their topic or their content is uncommon. In the alt-text, of course. Even though explanations in the alt-text are not accessible to people with certain disabilities. But most Mastodon users can't imagine that at least some Fediverse users can pour thousands or even millions of characters into the message text.
Nonetheless, this gives the impression of Mastodon's culture demanding any and all extra information that people may need about an image be immediately delivered on a silver platter, along with the image.
Granted, if what the image shows is obscure enough, and especially if it also might make people curious enough, this appears to be justified. The former definitely does apply to my own original images, and chances are the latter applies, too. Technically speaking, my original images need a whole lot of explanation because nobody has ever even heard of what they show, and they also need highly detailed descriptions because nobody who can't see the image has even only the faintest idea of what anything in the image looks like.
So what I do is describe each one of my original images twice. One description is the long description; it's fully detailed, it contains transcripts of all bits of text within the borders of the image, and it contains all necessary explanations. It goes into the message text, and it's pretty much always a five-digit number of characters long.
The other description is shorter, and it goes into the alt-text in order to satisfy those alt-text hardliners who demand there be an actually descriptive alt-text even if there's an extremely detailed image description in the message text. I used to fill Mastodon's limit of 1,500 characters up almost or actually entirely. My future alt-texts might be limited to a bit over 500 characters because Misskey and its many forks delete alt-texts that are longer than 512 characters. (Then again, they also discard anything with a message text of over 10,000 characters, I think.)
Describing one image takes me from several hours up to two full days. I rarely receive any praise for this effort. But I guess the downside is that each original image that I post has the channel on which I post my original images now (as rarely as this happens) silently muted or blocked by a whole number of Mastodon users for exceeding the holy 500-character limit.
I try to play it safe. I try to always be ahead of Mastodon's alt-text police, far enough for them not to demand I edit the image descriptions of a three-year-old image post because they're sub-standard. So in addition to these monster descriptions, I have over 30 rules and guidelines for describing images and writing alt-texts that I adhere to as strictly as I possibly can. And yes, I also expect most of them to be enforced by the alt-text police sooner or later.
CC: @Princess Nettles Succubish
#Long #LongPost #CWLong #CWLongPost #FediMeta #FediverseMeta #CWFediMeta #CWFediverseMeta #CharacterLimit #CharacterLimits #CharacterLimitMeta #CWCharacterLimitMeta #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta #MastodonCulture -
@✨pencilears✨(KD7POW)
I think a lot of people have unrealistically high expectations for alt text, and as a consequence they add a bunch of fluff, which makes writing alt-text feel daunting if that's the standard.
Looking at this phenomenon from outside of Mastodon (I'm on Hubzilla), I've got the impression that this has become engrained into Mastodon's very culture.
Mastodon has a limit of 1,500 characters so that the text in images with a lot of text can be transcribed in alt-text. However, many Mastodon users, knowing everything they know about online accessibility, take this as an invitation to write very detailed visual descriptions. Other Mastodon users cheer and applaud for these long descriptions.
Maybe it's just my impression. Maybe I'm wrong. But I kind of expect at least some Mastodon users to start both demanding and enforcing very long and detailed alt-texts/image descriptions, regardless of long alt-texts actually being bad for various reasons and even Mastodon users saying so.
I mean, I've already read several times that images must also be explained if their topic or their content is uncommon. In the alt-text, of course. Even though explanations in the alt-text are not accessible to people with certain disabilities. But most Mastodon users can't imagine that at least some Fediverse users can pour thousands or even millions of characters into the message text.
Nonetheless, this gives the impression of Mastodon's culture demanding any and all extra information that people may need about an image be immediately delivered on a silver platter, along with the image.
Granted, if what the image shows is obscure enough, and especially if it also might make people curious enough, this appears to be justified. The former definitely does apply to my own original images, and chances are the latter applies, too. Technically speaking, my original images need a whole lot of explanation because nobody has ever even heard of what they show, and they also need highly detailed descriptions because nobody who can't see the image has even only the faintest idea of what anything in the image looks like.
So what I do is describe each one of my original images twice. One description is the long description; it's fully detailed, it contains transcripts of all bits of text within the borders of the image, and it contains all necessary explanations. It goes into the message text, and it's pretty much always a five-digit number of characters long.
The other description is shorter, and it goes into the alt-text in order to satisfy those alt-text hardliners who demand there be an actually descriptive alt-text even if there's an extremely detailed image description in the message text. I used to fill Mastodon's limit of 1,500 characters up almost or actually entirely. My future alt-texts might be limited to a bit over 500 characters because Misskey and its many forks delete alt-texts that are longer than 512 characters. (Then again, they also discard anything with a message text of over 10,000 characters, I think.)
Describing one image takes me from several hours up to two full days. I rarely receive any praise for this effort. But I guess the downside is that each original image that I post has the channel on which I post my original images now (as rarely as this happens) silently muted or blocked by a whole number of Mastodon users for exceeding the holy 500-character limit.
I try to play it safe. I try to always be ahead of Mastodon's alt-text police, far enough for them not to demand I edit the image descriptions of a three-year-old image post because they're sub-standard. So in addition to these monster descriptions, I have over 30 rules and guidelines for describing images and writing alt-texts that I adhere to as strictly as I possibly can. And yes, I also expect most of them to be enforced by the alt-text police sooner or later.
CC: @Princess Nettles Succubish
#Long #LongPost #CWLong #CWLongPost #FediMeta #FediverseMeta #CWFediMeta #CWFediverseMeta #CharacterLimit #CharacterLimits #CharacterLimitMeta #CWCharacterLimitMeta #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta #MastodonCulture -
@✨pencilears✨(KD7POW)
I think a lot of people have unrealistically high expectations for alt text, and as a consequence they add a bunch of fluff, which makes writing alt-text feel daunting if that's the standard.
Looking at this phenomenon from outside of Mastodon (I'm on Hubzilla), I've got the impression that this has become engrained into Mastodon's very culture.
Mastodon has a limit of 1,500 characters so that the text in images with a lot of text can be transcribed in alt-text. However, many Mastodon users, knowing everything they know about online accessibility, take this as an invitation to write very detailed visual descriptions. Other Mastodon users cheer and applaud for these long descriptions.
Maybe it's just my impression. Maybe I'm wrong. But I kind of expect at least some Mastodon users to start both demanding and enforcing very long and detailed alt-texts/image descriptions, regardless of long alt-texts actually being bad for various reasons and even Mastodon users saying so.
I mean, I've already read several times that images must also be explained if their topic or their content is uncommon. In the alt-text, of course. Even though explanations in the alt-text are not accessible to people with certain disabilities. But most Mastodon users can't imagine that at least some Fediverse users can pour thousands or even millions of characters into the message text.
Nonetheless, this gives the impression of Mastodon's culture demanding any and all extra information that people may need about an image be immediately delivered on a silver platter, along with the image.
Granted, if what the image shows is obscure enough, and especially if it also might make people curious enough, this appears to be justified. The former definitely does apply to my own original images, and chances are the latter applies, too. Technically speaking, my original images need a whole lot of explanation because nobody has ever even heard of what they show, and they also need highly detailed descriptions because nobody who can't see the image has even only the faintest idea of what anything in the image looks like.
So what I do is describe each one of my original images twice. One description is the long description; it's fully detailed, it contains transcripts of all bits of text within the borders of the image, and it contains all necessary explanations. It goes into the message text, and it's pretty much always a five-digit number of characters long.
The other description is shorter, and it goes into the alt-text in order to satisfy those alt-text hardliners who demand there be an actually descriptive alt-text even if there's an extremely detailed image description in the message text. I used to fill Mastodon's limit of 1,500 characters up almost or actually entirely. My future alt-texts might be limited to a bit over 500 characters because Misskey and its many forks delete alt-texts that are longer than 512 characters. (Then again, they also discard anything with a message text of over 10,000 characters, I think.)
Describing one image takes me from several hours up to two full days. I rarely receive any praise for this effort. But I guess the downside is that each original image that I post has the channel on which I post my original images now (as rarely as this happens) silently muted or blocked by a whole number of Mastodon users for exceeding the holy 500-character limit.
I try to play it safe. I try to always be ahead of Mastodon's alt-text police, far enough for them not to demand I edit the image descriptions of a three-year-old image post because they're sub-standard. So in addition to these monster descriptions, I have over 30 rules and guidelines for describing images and writing alt-texts that I adhere to as strictly as I possibly can. And yes, I also expect most of them to be enforced by the alt-text police sooner or later.
CC: @Princess Nettles Succubish
#Long #LongPost #CWLong #CWLongPost #FediMeta #FediverseMeta #CWFediMeta #CWFediverseMeta #CharacterLimit #CharacterLimits #CharacterLimitMeta #CWCharacterLimitMeta #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta #MastodonCulture -
@✨pencilears✨(KD7POW)I think a lot of people have unrealistically high expectations for alt text, and as a consequence they add a bunch of fluff, which makes writing alt-text feel daunting if that's the standard.
Looking at this phenomenon from outside of Mastodon (I'm on Hubzilla), I've got the impression that this has become engrained into Mastodon's very culture.
Mastodon has a limit of 1,500 characters so that the text in images with a lot of text can be transcribed in alt-text. However, many Mastodon users, knowing everything they know about online accessibility, take this as an invitation to write very detailed visual descriptions. Other Mastodon users cheer and applaud for these long descriptions.
Maybe it's just my impression. Maybe I'm wrong. But I kind of expect at least some Mastodon users to start both demanding and enforcing very long and detailed alt-texts/image descriptions, regardless of long alt-texts actually being bad for various reasons and even Mastodon users saying so.
I mean, I've already read several times that images must also be explained if their topic or their content is uncommon. In the alt-text, of course. Even though explanations in the alt-text are not accessible to people with certain disabilities. But most Mastodon users can't imagine that at least some Fediverse users can pour thousands or even millions of characters into the message text.
Nonetheless, this gives the impression of Mastodon's culture demanding any and all extra information that people may need about an image be immediately delivered on a silver platter, along with the image.
Granted, if what the image shows is obscure enough, and especially if it also might make people curious enough, this appears to be justified. The former definitely does apply to my own original images, and chances are the latter applies, too. Technically speaking, my original images need a whole lot of explanation because nobody has ever even heard of what they show, and they also need highly detailed descriptions because nobody who can't see the image has even only the faintest idea of what anything in the image looks like.
So what I do is describe each one of my original images twice. One description is the long description; it's fully detailed, it contains transcripts of all bits of text within the borders of the image, and it contains all necessary explanations. It goes into the message text, and it's pretty much always a five-digit number of characters long.
The other description is shorter, and it goes into the alt-text in order to satisfy those alt-text hardliners who demand there be an actually descriptive alt-text even if there's an extremely detailed image description in the message text. I used to fill Mastodon's limit of 1,500 characters up almost or actually entirely. My future alt-texts might be limited to a bit over 500 characters because Misskey and its many forks delete alt-texts that are longer than 512 characters. (Then again, they also discard anything with a message text of over 10,000 characters, I think.)
Describing one image takes me from several hours up to two full days. I rarely receive any praise for this effort. But I guess the downside is that each original image that I post has the channel on which I post my original images now (as rarely as this happens) silently muted or blocked by a whole number of Mastodon users for exceeding the holy 500-character limit.
I try to play it safe. I try to always be ahead of Mastodon's alt-text police, far enough for them not to demand I edit the image descriptions of a three-year-old image post because they're sub-standard. So in addition to these monster descriptions, I have over 30 rules and guidelines for describing images and writing alt-texts that I adhere to as strictly as I possibly can. And yes, I also expect most of them to be enforced by the alt-text police sooner or later.
CC: @Princess Nettles Succubish
#Long #LongPost #CWLong #CWLongPost #FediMeta #FediverseMeta #CWFediMeta #CWFediverseMeta #CharacterLimit #CharacterLimits #CharacterLimitMeta #CWCharacterLimitMeta #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta #MastodonCulture -
@Fardels Bear
An image is a complex thing, and alt text only tells you what the poster saw, which elements were important enough for them to highlight.
So you say it'd be better if all images were always described in full detail, regardless of context, just in case someone somewhere out there is interested in details in the image that don't matter in the context?
I do just this with my original images. I have enough room in my posts for literally novel-sized image descriptions, what with no real character limit. And for those who absolutely demand that each image have an image description in the alt-text, I distill another, shorter description from the long one and put it in the alt-text, only that this description isn't fully detailed.
However, describing one image takes me from several hours up to several days. I've once taken two full days, morning to evening, to describe one image in 60,000 characters. Plus another morning to write an extra description of 1,400 characters for the alt-text. This is also why I haven't posted a single original image in over two years.
When I post memes that aren't based on original images, I don't write fully detailed image descriptions. I limit them to what matters, and explaining the meme in the message text takes up enough time and energy already. But if that's bad...
On top of all this comes that there are dozens of rules and guidelines for describing images and writing alt-texts that must be followed, at least if you want to write actual quality image descriptions and alt-texts that work with all frontends, that don't offend anyone etc. I know them. I collect them because I'm working on a very extensive wiki about alt-texts and image descriptions for the Fediverse.
These rules and guidelines make the difference between writing some image descriptions/alt-texts and writing good image descriptions/alt-texts. Right now, nobody seems to enforce them and sanction those who don't follow them. But at least I want to be prepared for when this happens.
So much about writing image descriptions and alt-texts being easy. At least writing good image descriptions and alt-texts isn't.
Still, I don't really want to go back and edit all my image descriptions and alt-texts whenever I uncover another rule or another guideline to follow.
CC: @Roknrol
#Long #LongPost #CWLong #CWLongPost #CharacterLimit #CharacterLimits #CharacterLimitMeta #CWCharacterLimitMeta #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta -
@Fardels Bear
An image is a complex thing, and alt text only tells you what the poster saw, which elements were important enough for them to highlight.
So you say it'd be better if all images were always described in full detail, regardless of context, just in case someone somewhere out there is interested in details in the image that don't matter in the context?
I do just this with my original images. I have enough room in my posts for literally novel-sized image descriptions, what with no real character limit. And for those who absolutely demand that each image have an image description in the alt-text, I distill another, shorter description from the long one and put it in the alt-text, only that this description isn't fully detailed.
However, describing one image takes me from several hours up to several days. I've once taken two full days, morning to evening, to describe one image in 60,000 characters. Plus another morning to write an extra description of 1,400 characters for the alt-text. This is also why I haven't posted a single original image in over two years.
When I post memes that aren't based on original images, I don't write fully detailed image descriptions. I limit them to what matters, and explaining the meme in the message text takes up enough time and energy already. But if that's bad...
On top of all this comes that there are dozens of rules and guidelines for describing images and writing alt-texts that must be followed, at least if you want to write actual quality image descriptions and alt-texts that work with all frontends, that don't offend anyone etc. I know them. I collect them because I'm working on a very extensive wiki about alt-texts and image descriptions for the Fediverse.
These rules and guidelines make the difference between writing some image descriptions/alt-texts and writing good image descriptions/alt-texts. Right now, nobody seems to enforce them and sanction those who don't follow them. But at least I want to be prepared for when this happens.
So much about writing image descriptions and alt-texts being easy. At least writing good image descriptions and alt-texts isn't.
Still, I don't really want to go back and edit all my image descriptions and alt-texts whenever I uncover another rule or another guideline to follow.
CC: @Roknrol
#Long #LongPost #CWLong #CWLongPost #CharacterLimit #CharacterLimits #CharacterLimitMeta #CWCharacterLimitMeta #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta -
@Fardels Bear
An image is a complex thing, and alt text only tells you what the poster saw, which elements were important enough for them to highlight.
So you say it'd be better if all images were always described in full detail, regardless of context, just in case someone somewhere out there is interested in details in the image that don't matter in the context?
I do just this with my original images. I have enough room in my posts for literally novel-sized image descriptions, what with no real character limit. And for those who absolutely demand that each image have an image description in the alt-text, I distill another, shorter description from the long one and put it in the alt-text, only that this description isn't fully detailed.
However, describing one image takes me from several hours up to several days. I've once taken two full days, morning to evening, to describe one image in 60,000 characters. Plus another morning to write an extra description of 1,400 characters for the alt-text. This is also why I haven't posted a single original image in over two years.
When I post memes that aren't based on original images, I don't write fully detailed image descriptions. I limit them to what matters, and explaining the meme in the message text takes up enough time and energy already. But if that's bad...
On top of all this comes that there are dozens of rules and guidelines for describing images and writing alt-texts that must be followed, at least if you want to write actual quality image descriptions and alt-texts that work with all frontends, that don't offend anyone etc. I know them. I collect them because I'm working on a very extensive wiki about alt-texts and image descriptions for the Fediverse.
These rules and guidelines make the difference between writing some image descriptions/alt-texts and writing good image descriptions/alt-texts. Right now, nobody seems to enforce them and sanction those who don't follow them. But at least I want to be prepared for when this happens.
So much about writing image descriptions and alt-texts being easy. At least writing good image descriptions and alt-texts isn't.
Still, I don't really want to go back and edit all my image descriptions and alt-texts whenever I uncover another rule or another guideline to follow.
CC: @Roknrol
#Long #LongPost #CWLong #CWLongPost #CharacterLimit #CharacterLimits #CharacterLimitMeta #CWCharacterLimitMeta #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta -
@Fardels Bear
An image is a complex thing, and alt text only tells you what the poster saw, which elements were important enough for them to highlight.
So you say it'd be better if all images were always described in full detail, regardless of context, just in case someone somewhere out there is interested in details in the image that don't matter in the context?
I do just this with my original images. I have enough room in my posts for literally novel-sized image descriptions, what with no real character limit. And for those who absolutely demand that each image have an image description in the alt-text, I distill another, shorter description from the long one and put it in the alt-text, only that this description isn't fully detailed.
However, describing one image takes me from several hours up to several days. I've once taken two full days, morning to evening, to describe one image in 60,000 characters. Plus another morning to write an extra description of 1,400 characters for the alt-text. This is also why I haven't posted a single original image in over two years.
When I post memes that aren't based on original images, I don't write fully detailed image descriptions. I limit them to what matters, and explaining the meme in the message text takes up enough time and energy already. But if that's bad...
On top of all this comes that there are dozens of rules and guidelines for describing images and writing alt-texts that must be followed, at least if you want to write actual quality image descriptions and alt-texts that work with all frontends, that don't offend anyone etc. I know them. I collect them because I'm working on a very extensive wiki about alt-texts and image descriptions for the Fediverse.
These rules and guidelines make the difference between writing some image descriptions/alt-texts and writing good image descriptions/alt-texts. Right now, nobody seems to enforce them and sanction those who don't follow them. But at least I want to be prepared for when this happens.
So much about writing image descriptions and alt-texts being easy. At least writing good image descriptions and alt-texts isn't.
Still, I don't really want to go back and edit all my image descriptions and alt-texts whenever I uncover another rule or another guideline to follow.
CC: @Roknrol
#Long #LongPost #CWLong #CWLongPost #CharacterLimit #CharacterLimits #CharacterLimitMeta #CWCharacterLimitMeta #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta -
@Nicolas Mouart noPhD, noMA No. I'm someone who wants to write image descriptions and alt-text not somehow, but the best way I can.
For this, I've been gathering information from a whole lot of sources for more than three years. More recently, I've started compiling it in a wiki so I can share it with others. When the wiki is done, it will have over 30 individual rules and guidelines for describing images and writing alt-texts.
125 characters as a limit are widespread because screen readers used to have a 125-character hard cap at some point in the 2010s or so. They simply couldn't read alt-texts beyond the 125-character mark. A while later, the hard limit was raised to 200 characters.
Nowadays, there is no hard character limit in screen readers to speak of anymore. Mastodon user @Eric Eggert says so, too: https://yatil.net/blog/there-is-no-character-limit-for-alt-text
The lowest actual character limit that must be heeded in the Fediverse is 512 characters. It is imposed by Misskey, a microblogging server application that's part of the Fediverse and fully federated with Mastodon, but more powerful than Mastodon (https://misskey-hub.net/en/; https://en.wikipedia.org/wiki/Misskey), as well as its various forks and forks of forks such as Calckey (defunct, but servers still exist), Firefish (ditto), Iceshrimp-JS, Sharkey, Catodon or CherryPick.
By design, Misskey is supposed to truncate alt-texts that are longer than 512 characters at the 512-character mark. However, due to a bug that still isn't fixed, it deletes the alt-text instead. The same applies to all the "Forkeys". If you write a longer alt-text, users on Misskey and the Forkeys will believe you haven't supplied an alt-text to begin with. You take a big effort upon you and write this monster of an alt-text, and regardless, users on Misskey, Sharkey & Co. will think you're another ableist swine who can't be bothered.
The only other thing that speaks against the 1000+-character alt-texts that are so popular on Mastodon is simply useability: A screen reader can navigate normal text, i.e. "rewind" a few words or to the beginning of the sentence of whatever. It can't navigate alt-text in this way, though. So let's suppose you're having a 1000-character alt-text read out to you. Around the 900-character mark, something distracts you. You have to have the last sentence re-read. No dice. The only thing the screen reader can do is jump back to the beginning of the alt-text and re-read the whole thing.
Still, Mastodon, being as Mastodon-centric as it has always been, loves long alt-texts. For one, it loves to use alt-texts to write around the 500-character limit which, by the way, is very bad.
But it also appears to love long and detailed image descriptions. They appear to have become part of Mastodon's culture. I'm actually waiting for Fediverse users being sanctioned in some way for providing alt-texts that adhere to Web accessibility standards, but that are too short for at least some Mastodon users' standards. In fact, I'm waiting for the first Mastodon user sanctioning someone else because they consider an alt-text of 500 characters too short and lacking details.
#Long #LongPost #CWLong #CWLongPost #FediMeta #FediverseMeta #CWFediMeta #CWFediverseMeta #Fediverse #Mastodon #Misskey #Forkey #Forkeys #Calckey #Firefish #Iceshrimp #IceshrimpJS #Sharkey #Catodon #CherryPick #MastodonCulture #MastodonCentricity #MastodonNormativity #CharacterLimit #CharacterLimits #CharacterLimitMeta #CWCharacterLimitMeta #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta -
@Nicolas Mouart noPhD, noMA As for the character limit, that's because Gemini is sourcing old and outdated information of which it doesn't know that it's old and outdated.
I'm working on an extensive wiki about image descriptions and alt-texts for the Fediverse. This is the page about how long alt-text can and should be: https://hub.netzgemeinde.eu/wiki/jupiter_rowland/How(20)to(20)describe(20)images(20)in(20)the(20)Fediverse/How(20)long(20)can(20)alt-text(20)be(3f)
#Long #LongPost #CWLong #CWLongPost #FediMeta #FediverseMeta #CWFediMeta #CWFediverseMeta #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta -
@Mark Wyner Won’t Comply :vm: First of all, you completely forgot the social media posters. The TOC link leads nowhere.
Also, maybe it's fair to mention that I'm working on an entire wiki about alt-texts and image descriptions, especially for the Fediverse. 46 pages are planned as of now, 24 of them are already written. These will include over 30 pages on how to do things, how not to do things and what never to do at all.
I won't add step-by-step guides for certain frontends. But I've already started adding rules and guidelines for how to describe images appropriately and write alt-texts in a way that won't break stuff anywhere.
Even if it's a very incomplete WIP, it's unique in several ways:- It's the only wiki about alt-texts and image descriptions. Few writers have even only written multiple pages or articles on this subject.
- It's the only guide that takes the existence of the Fediverse outside of Mastodon into consideration. This means that it's also the only guide that doesn't assume that Fediverse equals always having only 500 characters for the message text. It actually recommends an additional, longer image description in the message text in certain situations, regardless of whether or not vanilla Mastodon users can do that.
- It's the only guide that, instead of regurgitating the alt-text rules for webpages and blogs and assuming that they apply to the Fediverse just the same, talks about the peculiarities of posting images into the Fediverse, how Mastodon's alt-text quality criteria are different from the W3C WAI and WCAG alt-text quality criteria, and why it's critically important to know and fulfill Mastodon's requirements for good alt-texts.
- It's probably the only guide that doesn't assume that you can always precisely target a certain audience for your content and make sure that only those receive it who understand it right off the bat.
- It is the only guide that lists over 60 references, including your blog post, but also including university webpages as well as over two dozen guides from Veronica Lewis a.k.a. Veronica with Four Eyes.
#Long #LongPost #CWLong #CWLongPost #FediMeta #FediverseMeta #CWFediMeta #CWFediverseMeta #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta -
@Mark Wyner Won’t Comply :vm: First of all, you completely forgot the social media posters. The TOC link leads nowhere.
Also, maybe it's fair to mention that I'm working on an entire wiki about alt-texts and image descriptions, especially for the Fediverse. 46 pages are planned as of now, 24 of them are already written. These will include over 30 pages on how to do things, how not to do things and what never to do at all.
I won't add step-by-step guides for certain frontends. But I've already started adding rules and guidelines for how to describe images appropriately and write alt-texts in a way that won't break stuff anywhere.
Even if it's a very incomplete WIP, it's unique in several ways:- It's the only wiki about alt-texts and image descriptions. Few writers have even only written multiple pages or articles on this subject.
- It's the only guide that takes the existence of the Fediverse outside of Mastodon into consideration. This means that it's also the only guide that doesn't assume that Fediverse equals always having only 500 characters for the message text. It actually recommends an additional, longer image description in the message text in certain situations, regardless of whether or not vanilla Mastodon users can do that.
- It's the only guide that, instead of regurgitating the alt-text rules for webpages and blogs and assuming that they apply to the Fediverse just the same, talks about the peculiarities of posting images into the Fediverse, how Mastodon's alt-text quality criteria are different from the W3C WAI and WCAG alt-text quality criteria, and why it's critically important to know and fulfill Mastodon's requirements for good alt-texts.
- It's probably the only guide that doesn't assume that you can always precisely target a certain audience for your content and make sure that only those receive it who understand it right off the bat.
- It is the only guide that lists over 60 references, including your blog post, but also including university webpages as well as over two dozen guides from Veronica Lewis a.k.a. Veronica with Four Eyes.
#Long #LongPost #CWLong #CWLongPost #FediMeta #FediverseMeta #CWFediMeta #CWFediverseMeta #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta -
@Mark Wyner Won’t Comply :vm: First of all, you completely forgot the social media posters. The TOC link leads nowhere.
Also, maybe it's fair to mention that I'm working on an entire wiki about alt-texts and image descriptions, especially for the Fediverse. 46 pages are planned as of now, 24 of them are already written. These will include over 30 pages on how to do things, how not to do things and what never to do at all.
I won't add step-by-step guides for certain frontends. But I've already started adding rules and guidelines for how to describe images appropriately and write alt-texts in a way that won't break stuff anywhere.
Even if it's a very incomplete WIP, it's unique in several ways:- It's the only wiki about alt-texts and image descriptions. Few writers have even only written multiple pages or articles on this subject.
- It's the only guide that takes the existence of the Fediverse outside of Mastodon into consideration. This means that it's also the only guide that doesn't assume that Fediverse equals always having only 500 characters for the message text. It actually recommends an additional, longer image description in the message text in certain situations, regardless of whether or not vanilla Mastodon users can do that.
- It's the only guide that, instead of regurgitating the alt-text rules for webpages and blogs and assuming that they apply to the Fediverse just the same, talks about the peculiarities of posting images into the Fediverse, how Mastodon's alt-text quality criteria are different from the W3C WAI and WCAG alt-text quality criteria, and why it's critically important to know and fulfill Mastodon's requirements for good alt-texts.
- It's probably the only guide that doesn't assume that you can always precisely target a certain audience for your content and make sure that only those receive it who understand it right off the bat.
- It is the only guide that lists over 60 references, including your blog post, but also including university webpages as well as over two dozen guides from Veronica Lewis a.k.a. Veronica with Four Eyes.
#Long #LongPost #CWLong #CWLongPost #FediMeta #FediverseMeta #CWFediMeta #CWFediverseMeta #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta -
@Mark Wyner Won’t Comply :vm: First of all, you completely forgot the social media posters. The TOC link leads nowhere.
Also, maybe it's fair to mention that I'm working on an entire wiki about alt-texts and image descriptions, especially for the Fediverse. 46 pages are planned as of now, 24 of them are already written. These will include over 30 pages on how to do things, how not to do things and what never to do at all.
I won't add step-by-step guides for certain frontends. But I've already started adding rules and guidelines for how to describe images appropriately and write alt-texts in a way that won't break stuff anywhere.
Even if it's a very incomplete WIP, it's unique in several ways:- It's the only wiki about alt-texts and image descriptions. Few writers have even only written multiple pages or articles on this subject.
- It's the only guide that takes the existence of the Fediverse outside of Mastodon into consideration. This means that it's also the only guide that doesn't assume that Fediverse equals always having only 500 characters for the message text. It actually recommends an additional, longer image description in the message text in certain situations, regardless of whether or not vanilla Mastodon users can do that.
- It's the only guide that, instead of regurgitating the alt-text rules for webpages and blogs and assuming that they apply to the Fediverse just the same, talks about the peculiarities of posting images into the Fediverse, how Mastodon's alt-text quality criteria are different from the W3C WAI and WCAG alt-text quality criteria, and why it's critically important to know and fulfill Mastodon's requirements for good alt-texts.
- It's probably the only guide that doesn't assume that you can always precisely target a certain audience for your content and make sure that only those receive it who understand it right off the bat.
- It is the only guide that lists over 60 references, including your blog post, but also including university webpages as well as over two dozen guides from Veronica Lewis a.k.a. Veronica with Four Eyes.
#Long #LongPost #CWLong #CWLongPost #FediMeta #FediverseMeta #CWFediMeta #CWFediverseMeta #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta -
@On a Third Photography Write your image descriptions 100% by hand, full stop.
For starters, AI descriptions are bad. AIs hallucinate. AIs misidentify stuff if they don't know it. AIs don't mention stuff at all if they don't know it.
Besides, AIs can only generate generic image descriptions. They don't know the context. They describe stuff that doesn't matter for anyone. They completely omit stuff that does matter because they don't know that it matters. Only you yourself can individually tailor image descriptions to the context of the post.
#Long #LongPost #CWLong #CWLongPost #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta -
@On a Third Photography Write your image descriptions 100% by hand, full stop.
For starters, AI descriptions are bad. AIs hallucinate. AIs misidentify stuff if they don't know it. AIs don't mention stuff at all if they don't know it.
Besides, AIs can only generate generic image descriptions. They don't know the context. They describe stuff that doesn't matter for anyone. They completely omit stuff that does matter because they don't know that it matters. Only you yourself can individually tailor image descriptions to the context of the post.
#Long #LongPost #CWLong #CWLongPost #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta -
@On a Third Photography Write your image descriptions 100% by hand, full stop.
For starters, AI descriptions are bad. AIs hallucinate. AIs misidentify stuff if they don't know it. AIs don't mention stuff at all if they don't know it.
Besides, AIs can only generate generic image descriptions. They don't know the context. They describe stuff that doesn't matter for anyone. They completely omit stuff that does matter because they don't know that it matters. Only you yourself can individually tailor image descriptions to the context of the post.
#Long #LongPost #CWLong #CWLongPost #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta -
@On a Third Photography Write your image descriptions 100% by hand, full stop.
For starters, AI descriptions are bad. AIs hallucinate. AIs misidentify stuff if they don't know it. AIs don't mention stuff at all if they don't know it.
Besides, AIs can only generate generic image descriptions. They don't know the context. They describe stuff that doesn't matter for anyone. They completely omit stuff that does matter because they don't know that it matters. Only you yourself can individually tailor image descriptions to the context of the post.
#Long #LongPost #CWLong #CWLongPost #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta -
@Jonathan And that's the issue. It has to work with languages switching not only mid-text, but mid-sentence, like so (taken from a very old image description of mine):
The column between the passageways in the background to the left carries a black sign with group joiners for the Metropolis Newsnet group. The sign has the octagonal red Metropolis logo with a white M in the middle and a white edge around it in the top left corner. To the right of the logo, it is labelled in white, "Metropolis Newsnet Gruppe" which is partly German and translates to "Metropolis Newsnet Group". The writing below reads, "Aktuelle Informationen um Metropolis" and "Ankündigungen, Infos, Events, Fragen und Antworten im Chat" which is mostly German and translates to "Announcements, infos, events, questions and answers in the chat". The small neon green writing at the bottom edge reads, "Klick hier und folge dem Link im allgemeinen Chat, um der Gruppe beizutreten. Im Gruppenfenster JOIN klicken.(Click here to join the Metropolis Newsnet group." The German part of this translates to, "Click here and follow the link in the general chat to join the group. In the group window, click JOIN." Right below the black sign are three clickable white panels with black writings on them, "deutsch", "français" and "english".
Granted, I wouldn't put something this long into the alt-text. But I can't exclude someone being in a situation in which a text transcript that'll go into the alt-text causes the language to switch mid-sentence.
#Long #LongPost #CWLong #CWLongPost #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta #ForeignLanguage #Transcript #Transcripts -
@Jonathan And that's the issue. It has to work with languages switching not only mid-text, but mid-sentence, like so (taken from a very old image description of mine):
The column between the passageways in the background to the left carries a black sign with group joiners for the Metropolis Newsnet group. The sign has the octagonal red Metropolis logo with a white M in the middle and a white edge around it in the top left corner. To the right of the logo, it is labelled in white, "Metropolis Newsnet Gruppe" which is partly German and translates to "Metropolis Newsnet Group". The writing below reads, "Aktuelle Informationen um Metropolis" and "Ankündigungen, Infos, Events, Fragen und Antworten im Chat" which is mostly German and translates to "Announcements, infos, events, questions and answers in the chat". The small neon green writing at the bottom edge reads, "Klick hier und folge dem Link im allgemeinen Chat, um der Gruppe beizutreten. Im Gruppenfenster JOIN klicken.(Click here to join the Metropolis Newsnet group." The German part of this translates to, "Click here and follow the link in the general chat to join the group. In the group window, click JOIN." Right below the black sign are three clickable white panels with black writings on them, "deutsch", "français" and "english".
Granted, I wouldn't put something this long into the alt-text. But I can't exclude someone being in a situation in which a text transcript that'll go into the alt-text causes the language to switch mid-sentence.
#Long #LongPost #CWLong #CWLongPost #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta #ForeignLanguage #Transcript #Transcripts -
@Jonathan And that's the issue. It has to work with languages switching not only mid-text, but mid-sentence, like so (taken from a very old image description of mine):
The column between the passageways in the background to the left carries a black sign with group joiners for the Metropolis Newsnet group. The sign has the octagonal red Metropolis logo with a white M in the middle and a white edge around it in the top left corner. To the right of the logo, it is labelled in white, "Metropolis Newsnet Gruppe" which is partly German and translates to "Metropolis Newsnet Group". The writing below reads, "Aktuelle Informationen um Metropolis" and "Ankündigungen, Infos, Events, Fragen und Antworten im Chat" which is mostly German and translates to "Announcements, infos, events, questions and answers in the chat". The small neon green writing at the bottom edge reads, "Klick hier und folge dem Link im allgemeinen Chat, um der Gruppe beizutreten. Im Gruppenfenster JOIN klicken.(Click here to join the Metropolis Newsnet group." The German part of this translates to, "Click here and follow the link in the general chat to join the group. In the group window, click JOIN." Right below the black sign are three clickable white panels with black writings on them, "deutsch", "français" and "english".
Granted, I wouldn't put something this long into the alt-text. But I can't exclude someone being in a situation in which a text transcript that'll go into the alt-text causes the language to switch mid-sentence.
#Long #LongPost #CWLong #CWLongPost #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta #ForeignLanguage #Transcript #Transcripts -
@Jakob Rosin Come to think of it, ideally, I'd provide individual image descriptions with text transcripts in all languages that appear in the image myself.
This way, if I have a sign in French, English and German, I can add a verbatim transcript of the French text to the French description, a verbatim transcript of the English text to the English description and a verbatim transcript of the German text to the German description.
If I only provided an English description, and if I only gave translations of the French and German parts of the sign, neither French nor German readers would have a verbatim transcript of anything in the sign. French readers, for example, would get a French machine translation of my English translation of the French text instead of getting the French text itself as a verbatim transcript.
#Long #LongPost #CWLong #CWLongPost #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta #Transcript #Transcripts #ForeignLanguage -
@Jakob Rosin Come to think of it, ideally, I'd provide individual image descriptions with text transcripts in all languages that appear in the image myself.
This way, if I have a sign in French, English and German, I can add a verbatim transcript of the French text to the French description, a verbatim transcript of the English text to the English description and a verbatim transcript of the German text to the German description.
If I only provided an English description, and if I only gave translations of the French and German parts of the sign, neither French nor German readers would have a verbatim transcript of anything in the sign. French readers, for example, would get a French machine translation of my English translation of the French text instead of getting the French text itself as a verbatim transcript.
#Long #LongPost #CWLong #CWLongPost #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta #Transcript #Transcripts #ForeignLanguage -
@Jakob Rosin Come to think of it, ideally, I'd provide individual image descriptions with text transcripts in all languages that appear in the image myself.
This way, if I have a sign in French, English and German, I can add a verbatim transcript of the French text to the French description, a verbatim transcript of the English text to the English description and a verbatim transcript of the German text to the German description.
If I only provided an English description, and if I only gave translations of the French and German parts of the sign, neither French nor German readers would have a verbatim transcript of anything in the sign. French readers, for example, would get a French machine translation of my English translation of the French text instead of getting the French text itself as a verbatim transcript.
#Long #LongPost #CWLong #CWLongPost #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta #Transcript #Transcripts #ForeignLanguage -
@Jakob Rosin To be honest, what I had in mind wasn't screenshots of text although describing screenshots of foreign-language text can be tricky itself.
Here are some snippets from my first long image description. It's outdated in a way, but it provides good examples. And this image description did go into the message text (with an additional, shorter description in the alt-text).On the ramp to the teleporter, there is a black sign that reads, "Wenn Durchgehen nicht klappt, Klicken Sie das Bild zum Teleportieren. When Walk-Through does not work, Click the Image to teleport."
The column between the passageways in the background to the left carries a black sign with group joiners for the Metropolis Newsnet group. The sign has the octagonal red Metropolis logo with a white M in the middle and a white edge around it in the top left corner. To the right of the logo, it is labelled in white, "Metropolis Newsnet Gruppe" which is partly German and translates to "Metropolis Newsnet Group". The writing below reads, "Aktuelle Informationen um Metropolis" and "Ankündigungen, Infos, Events, Fragen und Antworten im Chat" which is mostly German and translates to "Announcements, infos, events, questions and answers in the chat". The small neon green writing at the bottom edge reads, "Klick hier und folge dem Link im allgemeinen Chat, um der Gruppe beizutreten. Im Gruppenfenster JOIN klicken.(Click here to join the Metropolis Newsnet group." The German part of this translates to, "Click here and follow the link in the general chat to join the group. In the group window, click JOIN." Right below the black sign are three clickable white panels with black writings on them, "deutsch", "français" and "english".
To my best knowledge, there's no way a screen reader can switch from English to German, to English, to German, to English within one sentence. Or from English to German, to French, back to English, again, within one sentence.
Or how about these two sentences from another, even longer image description?At the bottom of this area, "Patefacio radix" is written in medium dark grey letters, in a wide sans-serif typeface and in what is likely to be small caps. It is Latin for "open source".
The screen reader would have to switch from English to Latin, back to English, mid-sentence, not to mention that the screen reader would have to know how to properly pronounce a dead language.
#Long #LongPost #CWLong #CWLongPost #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta #Transcript #Transcripts #ForeignLanguage -
@Jakob Rosin To be honest, what I had in mind wasn't screenshots of text although describing screenshots of foreign-language text can be tricky itself.
Here are some snippets from my first long image description. It's outdated in a way, but it provides good examples. And this image description did go into the message text (with an additional, shorter description in the alt-text).On the ramp to the teleporter, there is a black sign that reads, "Wenn Durchgehen nicht klappt, Klicken Sie das Bild zum Teleportieren. When Walk-Through does not work, Click the Image to teleport."
The column between the passageways in the background to the left carries a black sign with group joiners for the Metropolis Newsnet group. The sign has the octagonal red Metropolis logo with a white M in the middle and a white edge around it in the top left corner. To the right of the logo, it is labelled in white, "Metropolis Newsnet Gruppe" which is partly German and translates to "Metropolis Newsnet Group". The writing below reads, "Aktuelle Informationen um Metropolis" and "Ankündigungen, Infos, Events, Fragen und Antworten im Chat" which is mostly German and translates to "Announcements, infos, events, questions and answers in the chat". The small neon green writing at the bottom edge reads, "Klick hier und folge dem Link im allgemeinen Chat, um der Gruppe beizutreten. Im Gruppenfenster JOIN klicken.(Click here to join the Metropolis Newsnet group." The German part of this translates to, "Click here and follow the link in the general chat to join the group. In the group window, click JOIN." Right below the black sign are three clickable white panels with black writings on them, "deutsch", "français" and "english".
To my best knowledge, there's no way a screen reader can switch from English to German, to English, to German, to English within one sentence. Or from English to German, to French, back to English, again, within one sentence.
Or how about these two sentences from another, even longer image description?At the bottom of this area, "Patefacio radix" is written in medium dark grey letters, in a wide sans-serif typeface and in what is likely to be small caps. It is Latin for "open source".
The screen reader would have to switch from English to Latin, back to English, mid-sentence, not to mention that the screen reader would have to know how to properly pronounce a dead language.
#Long #LongPost #CWLong #CWLongPost #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta #Transcript #Transcripts #ForeignLanguage -
@Jakob Rosin To be honest, what I had in mind wasn't screenshots of text although describing screenshots of foreign-language text can be tricky itself.
Here are some snippets from my first long image description. It's outdated in a way, but it provides good examples. And this image description did go into the message text (with an additional, shorter description in the alt-text).On the ramp to the teleporter, there is a black sign that reads, "Wenn Durchgehen nicht klappt, Klicken Sie das Bild zum Teleportieren. When Walk-Through does not work, Click the Image to teleport."
The column between the passageways in the background to the left carries a black sign with group joiners for the Metropolis Newsnet group. The sign has the octagonal red Metropolis logo with a white M in the middle and a white edge around it in the top left corner. To the right of the logo, it is labelled in white, "Metropolis Newsnet Gruppe" which is partly German and translates to "Metropolis Newsnet Group". The writing below reads, "Aktuelle Informationen um Metropolis" and "Ankündigungen, Infos, Events, Fragen und Antworten im Chat" which is mostly German and translates to "Announcements, infos, events, questions and answers in the chat". The small neon green writing at the bottom edge reads, "Klick hier und folge dem Link im allgemeinen Chat, um der Gruppe beizutreten. Im Gruppenfenster JOIN klicken.(Click here to join the Metropolis Newsnet group." The German part of this translates to, "Click here and follow the link in the general chat to join the group. In the group window, click JOIN." Right below the black sign are three clickable white panels with black writings on them, "deutsch", "français" and "english".
To my best knowledge, there's no way a screen reader can switch from English to German, to English, to German, to English within one sentence. Or from English to German, to French, back to English, again, within one sentence.
Or how about these two sentences from another, even longer image description?At the bottom of this area, "Patefacio radix" is written in medium dark grey letters, in a wide sans-serif typeface and in what is likely to be small caps. It is Latin for "open source".
The screen reader would have to switch from English to Latin, back to English, mid-sentence, not to mention that the screen reader would have to know how to properly pronounce a dead language.
#Long #LongPost #CWLong #CWLongPost #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta #Transcript #Transcripts #ForeignLanguage -
CW: How must foreign-language text be transcribed in an image description? CW: long (over 2,200 characters), alt-text meta, image description meta
I'm pretty sure that posting this will turn out completely futile. But I'll try anyway.
I'm especially asking the Mastodon alt-text enforcers: How must foreign-language text be transcribed, both by Web accessibility standards and by Mastodon's alt-text standards?
Let's assume I have some text in an image that's in a different language than the post and therefore in a different language than the image description. But I still have to transcribe it.
Now, according to image-describing rules, text must always be transcribed 100% verbatim. Word by word, character by character, exactly identical to the original in the image.
Here's the catch: Screen reader software may not be able to reliably switch the language in which they read out text in the middle of a description. They definitely cannot switch the language in the middle of an alt-text, I know that, just like they cannot navigate within alt-text.
This means that if I describe an image in English, but there's French text somewhere, a screen reader may read it out as gobbledygook instead of proper French. It may read the French text like it's English.
But even if the screen reader did read out the French text in proper French, someone who doesn't understand French won't understand the text.
This leaves me with two options:- Transcribe the foreign-language text verbatim, add a translation afterwards.
This would make the foreign-language text understandable, but it would also produce gobbledygook. - Give a translated transcript of the foreign-language text that's as close to the original as possible.
No gobbledygook, but I'd break the rule of always transcribing text verbatim.
The latter option would produce weird results if applied to a sign with essentially the same text in French, English and German where the French and German texts translate into something that slightly differs from both each other and the English text. I've happened upon this before.
#Long #LongPost #CWLong #CWLongPost #FediMeta #FediverseMeta #CWFediMeta #CWFediverseMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta #Transcript #Transcripts #ScreenReader #ScreenReaders #ForeignLanguage #Inclusion #A11y #Accessibility - Transcribe the foreign-language text verbatim, add a translation afterwards.
-
CW: How must foreign-language text be transcribed in an image description? CW: long (over 2,200 characters), alt-text meta, image description meta
I'm pretty sure that posting this will turn out completely futile. But I'll try anyway.
I'm especially asking the Mastodon alt-text enforcers: How must foreign-language text be transcribed, both by Web accessibility standards and by Mastodon's alt-text standards?
Let's assume I have some text in an image that's in a different language than the post and therefore in a different language than the image description. But I still have to transcribe it.
Now, according to image-describing rules, text must always be transcribed 100% verbatim. Word by word, character by character, exactly identical to the original in the image.
Here's the catch: Screen reader software may not be able to reliably switch the language in which they read out text in the middle of a description. They definitely cannot switch the language in the middle of an alt-text, I know that, just like they cannot navigate within alt-text.
This means that if I describe an image in English, but there's French text somewhere, a screen reader may read it out as gobbledygook instead of proper French. It may read the French text like it's English.
But even if the screen reader did read out the French text in proper French, someone who doesn't understand French won't understand the text.
This leaves me with two options:- Transcribe the foreign-language text verbatim, add a translation afterwards.
This would make the foreign-language text understandable, but it would also produce gobbledygook. - Give a translated transcript of the foreign-language text that's as close to the original as possible.
No gobbledygook, but I'd break the rule of always transcribing text verbatim.
The latter option would produce weird results if applied to a sign with essentially the same text in French, English and German where the French and German texts translate into something that slightly differs from both each other and the English text. I've happened upon this before.
#Long #LongPost #CWLong #CWLongPost #FediMeta #FediverseMeta #CWFediMeta #CWFediverseMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta #Transcript #Transcripts #ScreenReader #ScreenReaders #ForeignLanguage #Inclusion #A11y #Accessibility - Transcribe the foreign-language text verbatim, add a translation afterwards.
-
CW: How must foreign-language text be transcribed in an image description? CW: long (over 2,200 characters), alt-text meta, image description meta
I'm pretty sure that posting this will turn out completely futile. But I'll try anyway.
I'm especially asking the Mastodon alt-text enforcers: How must foreign-language text be transcribed, both by Web accessibility standards and by Mastodon's alt-text standards?
Let's assume I have some text in an image that's in a different language than the post and therefore in a different language than the image description. But I still have to transcribe it.
Now, according to image-describing rules, text must always be transcribed 100% verbatim. Word by word, character by character, exactly identical to the original in the image.
Here's the catch: Screen reader software may not be able to reliably switch the language in which they read out text in the middle of a description. They definitely cannot switch the language in the middle of an alt-text, I know that, just like they cannot navigate within alt-text.
This means that if I describe an image in English, but there's French text somewhere, a screen reader may read it out as gobbledygook instead of proper French. It may read the French text like it's English.
But even if the screen reader did read out the French text in proper French, someone who doesn't understand French won't understand the text.
This leaves me with two options:- Transcribe the foreign-language text verbatim, add a translation afterwards.
This would make the foreign-language text understandable, but it would also produce gobbledygook. - Give a translated transcript of the foreign-language text that's as close to the original as possible.
No gobbledygook, but I'd break the rule of always transcribing text verbatim.
The latter option would produce weird results if applied to a sign with essentially the same text in French, English and German where the French and German texts translate into something that slightly differs from both each other and the English text. I've happened upon this before.
#Long #LongPost #CWLong #CWLongPost #FediMeta #FediverseMeta #CWFediMeta #CWFediverseMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta #Transcript #Transcripts #ScreenReader #ScreenReaders #ForeignLanguage #Inclusion #A11y #Accessibility - Transcribe the foreign-language text verbatim, add a translation afterwards.
-
CW: How must foreign-language text be transcribed in an image description? CW: long (over 2,200 characters), alt-text meta, image description meta
I'm pretty sure that posting this will turn out completely futile. But I'll try anyway.
I'm especially asking the Mastodon alt-text enforcers: How must foreign-language text be transcribed, both by Web accessibility standards and by Mastodon's alt-text standards?
Let's assume I have some text in an image that's in a different language than the post and therefore in a different language than the image description. But I still have to transcribe it.
Now, according to image-describing rules, text must always be transcribed 100% verbatim. Word by word, character by character, exactly identical to the original in the image.
Here's the catch: Screen reader software may not be able to reliably switch the language in which they read out text in the middle of a description. They definitely cannot switch the language in the middle of an alt-text, I know that, just like they cannot navigate within alt-text.
This means that if I describe an image in English, but there's French text somewhere, a screen reader may read it out as gobbledygook instead of proper French. It may read the French text like it's English.
But even if the screen reader did read out the French text in proper French, someone who doesn't understand French won't understand the text.
This leaves me with two options:- Transcribe the foreign-language text verbatim, add a translation afterwards.
This would make the foreign-language text understandable, but it would also produce gobbledygook. - Give a translated transcript of the foreign-language text that's as close to the original as possible.
No gobbledygook, but I'd break the rule of always transcribing text verbatim.
The latter option would produce weird results if applied to a sign with essentially the same text in French, English and German where the French and German texts translate into something that slightly differs from both each other and the English text. I've happened upon this before.
#Long #LongPost #CWLong #CWLongPost #FediMeta #FediverseMeta #CWFediMeta #CWFediverseMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta #Transcript #Transcripts #ScreenReader #ScreenReaders #ForeignLanguage #Inclusion #A11y #Accessibility - Transcribe the foreign-language text verbatim, add a translation afterwards.
-
@Prof. Rachel Thorn 🍉🇺🇦🏳️⚧️🏳️ Just out of curiosity: Are there any good excuses for not having added an alt-text that you'd accept?
For example, when someone new to Hubzilla tells you that they neither know what alt-text is nor how to add it? Hubzilla doesn't have dedicated UI elements for adding alt-texts. Images are embedded into posts like in blog posts rather than attached as files like on Mastodon. Alt-texts have to be manually woven into the image-embedding phpBB code, and the newbie might not know that this is explained in Hubzilla's built-in help, or it's simply too tedious to "program" an alt-text in.
Or when a Hubzilla veteran of over ten years tells you that Hubzilla has no way whatsoever to add an alt-text to a video? Like images, videos are embedded into posts. But the video-embedding code has no option for an alt-text. Granted, if it's a video, I'd rather expect or supply a full, time-coded transcript in the message text. But maybe what's expected is both a full, time-coded transcript and an alt-text that's sufficiently detailed by Mastodon's standards.
If neither of these are acceptable to you, there's something else you should know: Hubzilla does not support Mastodon's report feature either because that's a non-standard, homegrown Mastodon feature, and Hubzilla is not built against Mastodon. So if you report someone to their Hubzilla admin, the admin simply will never receive your report.
This is likely to lead Mastodon users into believing that either the server on the other end is unmoderated (which, in Hubzilla's case and by Mastodon's standards, is technically true because Hubzilla is all about self-moderation), or the admin doesn't care, both of which, by Mastodon's standards, is reason enough to have the server Fediblocked.
Again, I'm just being curious and cautious. You should not believe that I'm trying to weasel myself out of having to describe images.
After all, I'm working on an extensive wiki about image descriptions and alt-texts in the Fediverse with 50+ pages with dozens of webpages and blog posts on how to describe images. I'm using the same knowledge that I'm pouring into the wiki, the same rules and guidelines that already have or will have wiki pages, to describe my own images, even though some of them collide with each other and with Mastodon's image description preferences and quality standards. I don't think anyone else in the Fediverse is taking describing images more seriously.
I rarely post images anymore due to the effort involved. But whenever I do, I take all the effort that's needed to both describe and explain them, taking into consideration the context, Mastodon's culture and preferences regarding alt-texts and image descriptions, the potential target audience and what knowledge they may or may not have about what's in the image. I've declared all my existing image descriptions outdated, also because it'd be too tedious to go and upgrade them all whenever I learn something new about describing images. (Don't go looking for my most recent images on this channel; I post my images elsewhere now because Hubzilla can't make Mastodon blur sensitive images.)
#Long #LongPost #CWLong #CWLongPost #FediMeta #FediverseMeta #CWFediMeta #CWFediverseMeta #Mastodon #Hubzilla #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta -
@Prof. Rachel Thorn 🍉🇺🇦🏳️⚧️🏳️ Just out of curiosity: Are there any good excuses for not having added an alt-text that you'd accept?
For example, when someone new to Hubzilla tells you that they neither know what alt-text is nor how to add it? Hubzilla doesn't have dedicated UI elements for adding alt-texts. Images are embedded into posts like in blog posts rather than attached as files like on Mastodon. Alt-texts have to be manually woven into the image-embedding phpBB code, and the newbie might not know that this is explained in Hubzilla's built-in help, or it's simply too tedious to "program" an alt-text in.
Or when a Hubzilla veteran of over ten years tells you that Hubzilla has no way whatsoever to add an alt-text to a video? Like images, videos are embedded into posts. But the video-embedding code has no option for an alt-text. Granted, if it's a video, I'd rather expect or supply a full, time-coded transcript in the message text. But maybe what's expected is both a full, time-coded transcript and an alt-text that's sufficiently detailed by Mastodon's standards.
If neither of these are acceptable to you, there's something else you should know: Hubzilla does not support Mastodon's report feature either because that's a non-standard, homegrown Mastodon feature, and Hubzilla is not built against Mastodon. So if you report someone to their Hubzilla admin, the admin simply will never receive your report.
This is likely to lead Mastodon users into believing that either the server on the other end is unmoderated (which, in Hubzilla's case and by Mastodon's standards, is technically true because Hubzilla is all about self-moderation), or the admin doesn't care, both of which, by Mastodon's standards, is reason enough to have the server Fediblocked.
Again, I'm just being curious and cautious. You should not believe that I'm trying to weasel myself out of having to describe images.
After all, I'm working on an extensive wiki about image descriptions and alt-texts in the Fediverse with 50+ pages with dozens of webpages and blog posts on how to describe images. I'm using the same knowledge that I'm pouring into the wiki, the same rules and guidelines that already have or will have wiki pages, to describe my own images, even though some of them collide with each other and with Mastodon's image description preferences and quality standards. I don't think anyone else in the Fediverse is taking describing images more seriously.
I rarely post images anymore due to the effort involved. But whenever I do, I take all the effort that's needed to both describe and explain them, taking into consideration the context, Mastodon's culture and preferences regarding alt-texts and image descriptions, the potential target audience and what knowledge they may or may not have about what's in the image. I've declared all my existing image descriptions outdated, also because it'd be too tedious to go and upgrade them all whenever I learn something new about describing images. (Don't go looking for my most recent images on this channel; I post my images elsewhere now because Hubzilla can't make Mastodon blur sensitive images.)
#Long #LongPost #CWLong #CWLongPost #FediMeta #FediverseMeta #CWFediMeta #CWFediverseMeta #Mastodon #Hubzilla #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta -
@Prof. Rachel Thorn 🍉🇺🇦🏳️⚧️🏳️ Just out of curiosity: Are there any good excuses for not having added an alt-text that you'd accept?
For example, when someone new to Hubzilla tells you that they neither know what alt-text is nor how to add it? Hubzilla doesn't have dedicated UI elements for adding alt-texts. Images are embedded into posts like in blog posts rather than attached as files like on Mastodon. Alt-texts have to be manually woven into the image-embedding phpBB code, and the newbie might not know that this is explained in Hubzilla's built-in help, or it's simply too tedious to "program" an alt-text in.
Or when a Hubzilla veteran of over ten years tells you that Hubzilla has no way whatsoever to add an alt-text to a video? Like images, videos are embedded into posts. But the video-embedding code has no option for an alt-text. Granted, if it's a video, I'd rather expect or supply a full, time-coded transcript in the message text. But maybe what's expected is both a full, time-coded transcript and an alt-text that's sufficiently detailed by Mastodon's standards.
If neither of these are acceptable to you, there's something else you should know: Hubzilla does not support Mastodon's report feature either because that's a non-standard, homegrown Mastodon feature, and Hubzilla is not built against Mastodon. So if you report someone to their Hubzilla admin, the admin simply will never receive your report.
This is likely to lead Mastodon users into believing that either the server on the other end is unmoderated (which, in Hubzilla's case and by Mastodon's standards, is technically true because Hubzilla is all about self-moderation), or the admin doesn't care, both of which, by Mastodon's standards, is reason enough to have the server Fediblocked.
Again, I'm just being curious and cautious. You should not believe that I'm trying to weasel myself out of having to describe images.
After all, I'm working on an extensive wiki about image descriptions and alt-texts in the Fediverse with 50+ pages with dozens of webpages and blog posts on how to describe images. I'm using the same knowledge that I'm pouring into the wiki, the same rules and guidelines that already have or will have wiki pages, to describe my own images, even though some of them collide with each other and with Mastodon's image description preferences and quality standards. I don't think anyone else in the Fediverse is taking describing images more seriously.
I rarely post images anymore due to the effort involved. But whenever I do, I take all the effort that's needed to both describe and explain them, taking into consideration the context, Mastodon's culture and preferences regarding alt-texts and image descriptions, the potential target audience and what knowledge they may or may not have about what's in the image. I've declared all my existing image descriptions outdated, also because it'd be too tedious to go and upgrade them all whenever I learn something new about describing images. (Don't go looking for my most recent images on this channel; I post my images elsewhere now because Hubzilla can't make Mastodon blur sensitive images.)
#Long #LongPost #CWLong #CWLongPost #FediMeta #FediverseMeta #CWFediMeta #CWFediverseMeta #Mastodon #Hubzilla #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta -
@OhSnap!Dragon You can type numbers. Screen readers will read them out as numbers and not as individual digits.
#AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta -
@OhSnap!Dragon You can type numbers. Screen readers will read them out as numbers and not as individual digits.
#AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta -
@OhSnap!Dragon You can type numbers. Screen readers will read them out as numbers and not as individual digits.
#AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta -
@OhSnap!Dragon You can type numbers. Screen readers will read them out as numbers and not as individual digits.
#AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta -
CW: Account getting away with way subpar alt-texts only because it's too niche with too few followers; CW: long (over 1,700 characters), Fediverse meta, alt-text meta, image description meta, character limit meta
I've just discovered a certain Mastodon account that seems to automatically post images of Second Life avatars. It can be lucky to only have eight followers, one of them being a search engine, the others being Second Life users.
If it had more follower, it would certainly already have met the wrath of the wider Mastodon community, especially the Mastodon HOA, for breaking Mastodon's unwritten rules.
And I'm not even talking about nigh-nudity in some of the images with no warning, no image flagging, no hashtag. I'm talking about the alt-texts that are vastly below Mastodon's requirements and quality standards for alt-texts.
Granted, all things considered, the requirements for good image descriptions by Mastodon's standards have to be extremely hard and tedious to meet, at least by my personal estimations. In fact, they have to be impossible to meet on Mastodon itself due to its tiny character limit, and if this account was somewhere where it could meet these requirements, it'd probably be blocked by loads of Mastodon accounts for its excessively long posts due to the long image descriptions.
But if more Mastodon users knew it right now, it'd probably be blocked by many more Mastodon accounts than follow it.
If you post from virtual worlds into the Fediverse, you simply cannot win in the long run. You'll lose either way.
#SecondLife #Metaverse #VirtualWorld #VirtualWorlds #Long #LongPost #CWLong #CWLongPost #FediMeta #FediverseMeta #CWFediMeta #CWFediverseMeta #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta #CharacterLimit #CharacterLimits #CharacterLimitMeta #CWCharacterLimitMeta #Mastodon #MastodonHOA #MastodonCulture -
CW: Account getting away with way subpar alt-texts only because it's too niche with too few followers; CW: long (over 1,700 characters), Fediverse meta, alt-text meta, image description meta, character limit meta
I've just discovered a certain Mastodon account that seems to automatically post images of Second Life avatars. It can be lucky to only have eight followers, one of them being a search engine, the others being Second Life users.
If it had more follower, it would certainly already have met the wrath of the wider Mastodon community, especially the Mastodon HOA, for breaking Mastodon's unwritten rules.
And I'm not even talking about nigh-nudity in some of the images with no warning, no image flagging, no hashtag. I'm talking about the alt-texts that are vastly below Mastodon's requirements and quality standards for alt-texts.
Granted, all things considered, the requirements for good image descriptions by Mastodon's standards have to be extremely hard and tedious to meet, at least by my personal estimations. In fact, they have to be impossible to meet on Mastodon itself due to its tiny character limit, and if this account was somewhere where it could meet these requirements, it'd probably be blocked by loads of Mastodon accounts for its excessively long posts due to the long image descriptions.
But if more Mastodon users knew it right now, it'd probably be blocked by many more Mastodon accounts than follow it.
If you post from virtual worlds into the Fediverse, you simply cannot win in the long run. You'll lose either way.
#SecondLife #Metaverse #VirtualWorld #VirtualWorlds #Long #LongPost #CWLong #CWLongPost #FediMeta #FediverseMeta #CWFediMeta #CWFediverseMeta #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta #CharacterLimit #CharacterLimits #CharacterLimitMeta #CWCharacterLimitMeta #Mastodon #MastodonHOA #MastodonCulture -
CW: Account getting away with way subpar alt-texts only because it's too niche with too few followers; CW: long (over 1,700 characters), Fediverse meta, alt-text meta, image description meta, character limit meta
I've just discovered a certain Mastodon account that seems to automatically post images of Second Life avatars. It can be lucky to only have eight followers, one of them being a search engine, the others being Second Life users.
If it had more follower, it would certainly already have met the wrath of the wider Mastodon community, especially the Mastodon HOA, for breaking Mastodon's unwritten rules.
And I'm not even talking about nigh-nudity in some of the images with no warning, no image flagging, no hashtag. I'm talking about the alt-texts that are vastly below Mastodon's requirements and quality standards for alt-texts.
Granted, all things considered, the requirements for good image descriptions by Mastodon's standards have to be extremely hard and tedious to meet, at least by my personal estimations. In fact, they have to be impossible to meet on Mastodon itself due to its tiny character limit, and if this account was somewhere where it could meet these requirements, it'd probably be blocked by loads of Mastodon accounts for its excessively long posts due to the long image descriptions.
But if more Mastodon users knew it right now, it'd probably be blocked by many more Mastodon accounts than follow it.
If you post from virtual worlds into the Fediverse, you simply cannot win in the long run. You'll lose either way.
#SecondLife #Metaverse #VirtualWorld #VirtualWorlds #Long #LongPost #CWLong #CWLongPost #FediMeta #FediverseMeta #CWFediMeta #CWFediverseMeta #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta #CharacterLimit #CharacterLimits #CharacterLimitMeta #CWCharacterLimitMeta #Mastodon #MastodonHOA #MastodonCulture -
CW: Account getting away with way subpar alt-texts only because it's too niche with too few followers; CW: long (over 1,700 characters), Fediverse meta, alt-text meta, image description meta, character limit meta
I've just discovered a certain Mastodon account that seems to automatically post images of Second Life avatars. It can be lucky to only have eight followers, one of them being a search engine, the others being Second Life users.
If it had more follower, it would certainly already have met the wrath of the wider Mastodon community, especially the Mastodon HOA, for breaking Mastodon's unwritten rules.
And I'm not even talking about nigh-nudity in some of the images with no warning, no image flagging, no hashtag. I'm talking about the alt-texts that are vastly below Mastodon's requirements and quality standards for alt-texts.
Granted, all things considered, the requirements for good image descriptions by Mastodon's standards have to be extremely hard and tedious to meet, at least by my personal estimations. In fact, they have to be impossible to meet on Mastodon itself due to its tiny character limit, and if this account was somewhere where it could meet these requirements, it'd probably be blocked by loads of Mastodon accounts for its excessively long posts due to the long image descriptions.
But if more Mastodon users knew it right now, it'd probably be blocked by many more Mastodon accounts than follow it.
If you post from virtual worlds into the Fediverse, you simply cannot win in the long run. You'll lose either way.
#SecondLife #Metaverse #VirtualWorld #VirtualWorlds #Long #LongPost #CWLong #CWLongPost #FediMeta #FediverseMeta #CWFediMeta #CWFediverseMeta #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta #CharacterLimit #CharacterLimits #CharacterLimitMeta #CWCharacterLimitMeta #Mastodon #MastodonHOA #MastodonCulture -
Quick #accessibility reminder for people who speak HTML: the #ImageDescription goes in the
altattribute, not undertitle.And yes, maybe some tools will work with both but others like the Copy Alt Text Firefox extension won't.
Found this thanks to XKCD where instead of the punchline I'm getting the comic title now. (Randall Munroe's site is famously inaccessible: go to explain xkcd instead, it has transcripts and shit.)
-
Quick #accessibility reminder for people who speak HTML: the #ImageDescription goes in the
altattribute, not undertitle.And yes, maybe some tools will work with both but others like the Copy Alt Text Firefox extension won't.
Found this thanks to XKCD where instead of the punchline I'm getting the comic title now. (Randall Munroe's site is famously inaccessible: go to explain xkcd instead, it has transcripts and shit.)
-
Quick #accessibility reminder for people who speak HTML: the #ImageDescription goes in the
altattribute, not undertitle.And yes, maybe some tools will work with both but others like the Copy Alt Text Firefox extension won't.
Found this thanks to XKCD where instead of the punchline I'm getting the comic title now. (Randall Munroe's site is famously inaccessible: go to explain xkcd instead, it has transcripts and shit.)
-
Quick #accessibility reminder for people who speak HTML: the #ImageDescription goes in the
altattribute, not undertitle.And yes, maybe some tools will work with both but others like the Copy Alt Text Firefox extension won't.
Found this thanks to XKCD where instead of the punchline I'm getting the comic title now. (Randall Munroe's site is famously inaccessible: go to explain xkcd instead, it has transcripts and shit.)
-
@Tyler Sticka @Old Man Vandehey @Neatnik :prami: I have a problem with this. But I probably only have this problem because whenever I post an image, it's about an extremely obscure niche topic. (Don't go looking for my recent images on this channel; I post my images elsewhere and even there only recently.)
When you're describing an image to a friend, first of all, you're describing the image to only one person. You probably already know that person. You can at least estimate what they know about the image you're going to describe. If not, you can ask them what they know, what they don't know, what they need to know, what they want to know.
And then you can tailor your description to the individual wants and needs of that one specific person. And you can add the explanations which this person needs and omit the explanations which this person doesn't need.
In case they don't understand something, or if they want to know more details, they can interrupt you mid-describing and ask you.
But when you're describing an image for the Fediverse, you're describing it for potentionally everyone in the Fediverse (and everyone on the Web, unless you restrict the permission to view your post). You don't know all these people, there's no way you can.
You simply can't go around and ask millions of Fediverse users and billions of Web users, one by one, what they know about the topic of your image, what they don't know etc. Neither can you predict who exactly will receive your post and then only go and ask these people in advance.
Even if you could, you can't tailor your image description and accompanying explanations to the individual needs of many people. Especially not if you're going to post that image only once for everyone all the same.
Worse yet, they can't interrupt you mid-describing and ask you for explanations either. They'll get the entire description plus your explanations all at once.
Thus, you have to describe and explain your image for everyone all the same with no regards for individual needs. You have to cover everyone's potential needs right away.
Let's take (link CW: eye contact, meme, non-Mastodon Fediverse meta) this image, for example. I've actually used it in a Fediverse meme post.
If I were to describe it to one friend, I'd ask the friend if they're familiar with the "One Does Not Simply Walk Into Mordor" meme template, and if they're familiar with FEP-ef61 "Portable Objects". Honestly, I expect hardly anyone to know both.
If they aren't familiar with FEP-ef61, and they most likely aren't, I'd ask them if they know nomadic identity, and if they know the streams repository, because knowing both is required for my explanation of FEP-ef61. They probably don't know either, so I'd explain the streams repository to them, including rattling down the whole 16-year history of the software family from Mistpark/Friendica to Forte. When I reach the rewriting of Red in 2012, I'd explain the concept of nomadic identity because it was then when it was first implemented. And even that I could only do if they knew the Fediverse. If not, I'd go start with the Fediverse.
If they aren't familiar with "One Does Not Simply Walk Into Mordor", I'd ask them if I could send them a link to the KnowYourMeme page, or if they need it described by me. If the latter, I'd ask them if they're familiar with the concepts of snowclones and image macros. And so forth. Whatever they aren't familiar with, I'd explain it to them.
As for the visuals, I'd ask them if they need a visual description of what Boromir looks like, regardless of whether or not this matters in the context. I'd also ask them how familiar they are with The Lord of the Rings and specifically the Peter Jackson films. For this may require even more explanations.
How am I supposed to do any of this with an audience of many? Right, I can't.
And so I have to deliver all necessary explanations in the message text, right beneath the image itself, just in case someone needs them. And let's face it, most people will need them. The explanation of the image macro itself plus another eight explanations in order for people to understand the main explanation. Over 25,000 characters if they consider links to external explanations bad.
#Long #LongPost #CWLong #CWLongPost #FediMeta #FediverseMeta #CWFediMeta #CWFediverseMeta #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta -
@Tyler Sticka @Old Man Vandehey @Neatnik :prami: I have a problem with this. But I probably only have this problem because whenever I post an image, it's about an extremely obscure niche topic. (Don't go looking for my recent images on this channel; I post my images elsewhere and even there only recently.)
When you're describing an image to a friend, first of all, you're describing the image to only one person. You probably already know that person. You can at least estimate what they know about the image you're going to describe. If not, you can ask them what they know, what they don't know, what they need to know, what they want to know.
And then you can tailor your description to the individual wants and needs of that one specific person. And you can add the explanations which this person needs and omit the explanations which this person doesn't need.
In case they don't understand something, or if they want to know more details, they can interrupt you mid-describing and ask you.
But when you're describing an image for the Fediverse, you're describing it for potentionally everyone in the Fediverse (and everyone on the Web, unless you restrict the permission to view your post). You don't know all these people, there's no way you can.
You simply can't go around and ask millions of Fediverse users and billions of Web users, one by one, what they know about the topic of your image, what they don't know etc. Neither can you predict who exactly will receive your post and then only go and ask these people in advance.
Even if you could, you can't tailor your image description and accompanying explanations to the individual needs of many people. Especially not if you're going to post that image only once for everyone all the same.
Worse yet, they can't interrupt you mid-describing and ask you for explanations either. They'll get the entire description plus your explanations all at once.
Thus, you have to describe and explain your image for everyone all the same with no regards for individual needs. You have to cover everyone's potential needs right away.
Let's take (link CW: eye contact, meme, non-Mastodon Fediverse meta) this image, for example. I've actually used it in a Fediverse meme post.
If I were to describe it to one friend, I'd ask the friend if they're familiar with the "One Does Not Simply Walk Into Mordor" meme template, and if they're familiar with FEP-ef61 "Portable Objects". Honestly, I expect hardly anyone to know both.
If they aren't familiar with FEP-ef61, and they most likely aren't, I'd ask them if they know nomadic identity, and if they know the streams repository, because knowing both is required for my explanation of FEP-ef61. They probably don't know either, so I'd explain the streams repository to them, including rattling down the whole 16-year history of the software family from Mistpark/Friendica to Forte. When I reach the rewriting of Red in 2012, I'd explain the concept of nomadic identity because it was then when it was first implemented. And even that I could only do if they knew the Fediverse. If not, I'd go start with the Fediverse.
If they aren't familiar with "One Does Not Simply Walk Into Mordor", I'd ask them if I could send them a link to the KnowYourMeme page, or if they need it described by me. If the latter, I'd ask them if they're familiar with the concepts of snowclones and image macros. And so forth. Whatever they aren't familiar with, I'd explain it to them.
As for the visuals, I'd ask them if they need a visual description of what Boromir looks like, regardless of whether or not this matters in the context. I'd also ask them how familiar they are with The Lord of the Rings and specifically the Peter Jackson films. For this may require even more explanations.
How am I supposed to do any of this with an audience of many? Right, I can't.
And so I have to deliver all necessary explanations in the message text, right beneath the image itself, just in case someone needs them. And let's face it, most people will need them. The explanation of the image macro itself plus another eight explanations in order for people to understand the main explanation. Over 25,000 characters if they consider links to external explanations bad.
#Long #LongPost #CWLong #CWLongPost #FediMeta #FediverseMeta #CWFediMeta #CWFediverseMeta #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta -
@Kiloku - Secretário do Caos @Neatnik :prami: I've been working on an entire wiki about image descriptions and alt-texts for the Fediverse for quite a while now. It's far from finished, not even half of the over 50 pages planned right now are written.
As of now, it'll have 32 rules and guidelines on what to do, how to do it and what not to do. The huge majority of these are completely unknown in the Fediverse otherwise.
#Long #LongPost #CWLong #CWLongPost #FediMeta #FediverseMeta #CWFediMeta #CWFediverseMeta #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta -
@Kiloku - Secretário do Caos @Neatnik :prami: I've been working on an entire wiki about image descriptions and alt-texts for the Fediverse for quite a while now. It's far from finished, not even half of the over 50 pages planned right now are written.
As of now, it'll have 32 rules and guidelines on what to do, how to do it and what not to do. The huge majority of these are completely unknown in the Fediverse otherwise.
#Long #LongPost #CWLong #CWLongPost #FediMeta #FediverseMeta #CWFediMeta #CWFediverseMeta #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta -
@Robert Kingett I've read somewhere that some blind people prefer LLM-generated image descriptions to human-written image descriptions because they're more entertaining. LLM-generated image descriptions often add some whimsy whereas human-written image descriptions just dryly rattle down what's in the image. Basically, through the screen reader, humans sound more like machines than actual machines.
They usually are aware that LLMs tend to be hallucinating and essentially telling them non-sense. But I've read from one blind user that they don't care whether or not the LLM-generated description is accurate as long as the accuracy isn't a matter of life and death.
In a certain way, it is understandable. The only people who'd criticise an image description for being inaccurate are fully sighted and therefore capable of comparing the image with its description.
Blind people won't notice unless the image description describes something so outlandish to them that they have the impression of an utterly surrealist image where there shouldn't be a surrealist image.
On the other hand, there are also blind people who demand image descriptions be accurate. They simply don't want to be told non-sense, especially not without knowing that they're being told non-sense.
But even on the sighted side, there's the "human versus AI" debate.
Some sighted people are fully convinced that LLMs can describe absolutely every image perfectly in absolutely every situation, no matter how obscure the contents of the image are. They're fully convinced that a generic LLM like ChatGPT can write circles around even human experts at any given time.
Some are simply AI fanbois or fangurls. Others say so in order to convince themselves that what they're doing is the best way: They use image-describing LLMs or even general-purpose LLMs as fire-and-forget tools. They have image descriptions generated, they copy-paste these image descriptions into the alt-texts, they send their posts, and they never take a look at these image descriptions at any point in the process. It's more convenient this way.
And then they wonder why they're under attack from sighted alt-text activists who call them out for their painfully inaccurate and blatantly obvious AI slop.
LLM proponents, both sighted and non-sighted, have in common that they never compare the image and the description. Non-sighted LLM proponents simply can't see the image. Sighted LLM proponents put so much faith into LLMs that they can't be bothered to read the description and cross-check it with the image.
#Long #LongPost #CWLong #CWLongPost #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta #AI #LLM #LLMs #AIVsHuman #HumanVsAI -
@Robert Kingett I've read somewhere that some blind people prefer LLM-generated image descriptions to human-written image descriptions because they're more entertaining. LLM-generated image descriptions often add some whimsy whereas human-written image descriptions just dryly rattle down what's in the image. Basically, through the screen reader, humans sound more like machines than actual machines.
They usually are aware that LLMs tend to be hallucinating and essentially telling them non-sense. But I've read from one blind user that they don't care whether or not the LLM-generated description is accurate as long as the accuracy isn't a matter of life and death.
In a certain way, it is understandable. The only people who'd criticise an image description for being inaccurate are fully sighted and therefore capable of comparing the image with its description.
Blind people won't notice unless the image description describes something so outlandish to them that they have the impression of an utterly surrealist image where there shouldn't be a surrealist image.
On the other hand, there are also blind people who demand image descriptions be accurate. They simply don't want to be told non-sense, especially not without knowing that they're being told non-sense.
But even on the sighted side, there's the "human versus AI" debate.
Some sighted people are fully convinced that LLMs can describe absolutely every image perfectly in absolutely every situation, no matter how obscure the contents of the image are. They're fully convinced that a generic LLM like ChatGPT can write circles around even human experts at any given time.
Some are simply AI fanbois or fangurls. Others say so in order to convince themselves that what they're doing is the best way: They use image-describing LLMs or even general-purpose LLMs as fire-and-forget tools. They have image descriptions generated, they copy-paste these image descriptions into the alt-texts, they send their posts, and they never take a look at these image descriptions at any point in the process. It's more convenient this way.
And then they wonder why they're under attack from sighted alt-text activists who call them out for their painfully inaccurate and blatantly obvious AI slop.
LLM proponents, both sighted and non-sighted, have in common that they never compare the image and the description. Non-sighted LLM proponents simply can't see the image. Sighted LLM proponents put so much faith into LLMs that they can't be bothered to read the description and cross-check it with the image.
#Long #LongPost #CWLong #CWLongPost #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta #AI #LLM #LLMs #AIVsHuman #HumanVsAI -
@Robert Kingett I've read somewhere that some blind people prefer LLM-generated image descriptions to human-written image descriptions because they're more entertaining. LLM-generated image descriptions often add some whimsy whereas human-written image descriptions just dryly rattle down what's in the image. Basically, through the screen reader, humans sound more like machines than actual machines.
They usually are aware that LLMs tend to be hallucinating and essentially telling them non-sense. But I've read from one blind user that they don't care whether or not the LLM-generated description is accurate as long as the accuracy isn't a matter of life and death.
In a certain way, it is understandable. The only people who'd criticise an image description for being inaccurate are fully sighted and therefore capable of comparing the image with its description.
Blind people won't notice unless the image description describes something so outlandish to them that they have the impression of an utterly surrealist image where there shouldn't be a surrealist image.
On the other hand, there are also blind people who demand image descriptions be accurate. They simply don't want to be told non-sense, especially not without knowing that they're being told non-sense.
But even on the sighted side, there's the "human versus AI" debate.
Some sighted people are fully convinced that LLMs can describe absolutely every image perfectly in absolutely every situation, no matter how obscure the contents of the image are. They're fully convinced that a generic LLM like ChatGPT can write circles around even human experts at any given time.
Some are simply AI fanbois or fangurls. Others say so in order to convince themselves that what they're doing is the best way: They use image-describing LLMs or even general-purpose LLMs as fire-and-forget tools. They have image descriptions generated, they copy-paste these image descriptions into the alt-texts, they send their posts, and they never take a look at these image descriptions at any point in the process. It's more convenient this way.
And then they wonder why they're under attack from sighted alt-text activists who call them out for their painfully inaccurate and blatantly obvious AI slop.
LLM proponents, both sighted and non-sighted, have in common that they never compare the image and the description. Non-sighted LLM proponents simply can't see the image. Sighted LLM proponents put so much faith into LLMs that they can't be bothered to read the description and cross-check it with the image.
#Long #LongPost #CWLong #CWLongPost #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta #AI #LLM #LLMs #AIVsHuman #HumanVsAI -
@Robert Kingett I've read somewhere that some blind people prefer LLM-generated image descriptions to human-written image descriptions because they're more entertaining. LLM-generated image descriptions often add some whimsy whereas human-written image descriptions just dryly rattle down what's in the image. Basically, through the screen reader, humans sound more like machines than actual machines.
They usually are aware that LLMs tend to be hallucinating and essentially telling them non-sense. But I've read from one blind user that they don't care whether or not the LLM-generated description is accurate as long as the accuracy isn't a matter of life and death.
In a certain way, it is understandable. The only people who'd criticise an image description for being inaccurate are fully sighted and therefore capable of comparing the image with its description.
Blind people won't notice unless the image description describes something so outlandish to them that they have the impression of an utterly surrealist image where there shouldn't be a surrealist image.
On the other hand, there are also blind people who demand image descriptions be accurate. They simply don't want to be told non-sense, especially not without knowing that they're being told non-sense.
But even on the sighted side, there's the "human versus AI" debate.
Some sighted people are fully convinced that LLMs can describe absolutely every image perfectly in absolutely every situation, no matter how obscure the contents of the image are. They're fully convinced that a generic LLM like ChatGPT can write circles around even human experts at any given time.
Some are simply AI fanbois or fangurls. Others say so in order to convince themselves that what they're doing is the best way: They use image-describing LLMs or even general-purpose LLMs as fire-and-forget tools. They have image descriptions generated, they copy-paste these image descriptions into the alt-texts, they send their posts, and they never take a look at these image descriptions at any point in the process. It's more convenient this way.
And then they wonder why they're under attack from sighted alt-text activists who call them out for their painfully inaccurate and blatantly obvious AI slop.
LLM proponents, both sighted and non-sighted, have in common that they never compare the image and the description. Non-sighted LLM proponents simply can't see the image. Sighted LLM proponents put so much faith into LLMs that they can't be bothered to read the description and cross-check it with the image.
#Long #LongPost #CWLong #CWLongPost #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta #AI #LLM #LLMs #AIVsHuman #HumanVsAI -
@DavidM_yeg @Dennis What more is needed?
A full visual description of the settings, the people (of course, avoiding their gender and describing their skin tone rather than their race), their outfits, the surroundings?
The title of the film, including an explanation what this film is, who made it, what it is about, when it was released, who the characters are and who the actors are?
Regardless of neither of this belonging into the alt-text because information that is neither available in the message text nor plainly visible to sighted users in the image does not belong into the alt-text? Because there are people who cannot access alt-text?
https://hub.netzgemeinde.eu/wiki/jupiter_rowland/How(20)to(20)describe(20)images(20)in(20)the(20)Fediverse/Don(27)t(20)explain(20)things(20)or(20)give(20)other(20)information(20)only(20)in(20)alt-text(21)
https://hub.netzgemeinde.eu/wiki/jupiter_rowland/How(20)to(20)describe(20)images(20)in(20)the(20)Fediverse/Can(20)everyone(20)access(20)alt-text(3f)
And, pray tell, how is all this supposed to fit in 512 characters or fewer so that Misskey, Calckey, Firefish, Iceshrimp-JS, Sharkey, CherryPick and the other Misskey forks won't discard the alt-text?
https://hub.netzgemeinde.eu/wiki/jupiter_rowland/How(20)to(20)describe(20)images(20)in(20)the(20)Fediverse/How(20)long(20)can(20)alt-text(20)be(3f)
#Long #LongPost #CWLong #CWLongPost #FediMeta #FediverseMeta #CWFediMeta #CWFediverseMeta #Misskey #Forkey #Forkeys #Calckey #Firefish #Iceshrimp #Iceshrimp-JS #Sharkey #CherryPick #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta -
@DavidM_yeg @Dennis What more is needed?
A full visual description of the settings, the people (of course, avoiding their gender and describing their skin tone rather than their race), their outfits, the surroundings?
The title of the film, including an explanation what this film is, who made it, what it is about, when it was released, who the characters are and who the actors are?
Regardless of neither of this belonging into the alt-text because information that is neither available in the message text nor plainly visible to sighted users in the image does not belong into the alt-text? Because there are people who cannot access alt-text?
https://hub.netzgemeinde.eu/wiki/jupiter_rowland/How(20)to(20)describe(20)images(20)in(20)the(20)Fediverse/Don(27)t(20)explain(20)things(20)or(20)give(20)other(20)information(20)only(20)in(20)alt-text(21)
https://hub.netzgemeinde.eu/wiki/jupiter_rowland/How(20)to(20)describe(20)images(20)in(20)the(20)Fediverse/Can(20)everyone(20)access(20)alt-text(3f)
And, pray tell, how is all this supposed to fit in 512 characters or fewer so that Misskey, Calckey, Firefish, Iceshrimp-JS, Sharkey, CherryPick and the other Misskey forks won't discard the alt-text?
https://hub.netzgemeinde.eu/wiki/jupiter_rowland/How(20)to(20)describe(20)images(20)in(20)the(20)Fediverse/How(20)long(20)can(20)alt-text(20)be(3f)
#Long #LongPost #CWLong #CWLongPost #FediMeta #FediverseMeta #CWFediMeta #CWFediverseMeta #Misskey #Forkey #Forkeys #Calckey #Firefish #Iceshrimp #Iceshrimp-JS #Sharkey #CherryPick #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta