-

2025-04-20 00:14:17

-0--1-

_9CL7T9k8cjnD_@mastodon.social

@iDGS "P.S. Always read the ALT text! 😘" I want to bear #AltBots Children. That being said, I wish the developers would find a way for AltBot to write to the AltTxt screen rather than tell me I need to be doing its AMAZINGLY GOOD and occasionally humorous job

show all 10 comments

2025-04-20 00:16:47

𝓐𝓷𝓭𝔂𝓣𝓲𝓮𝓭𝔂𝓮 𓀤

andytiedye@sfba.social

It would be nice to be be able to put alt text into the metadata of an image, so it populates automatically when the image is reshared. There are at least a couple of ways to do this, but none that Mastodon can recognize.

#AltText #Accessability

@_9CL7T9k8cjnD_ @iDGS

2025-04-20 17:29:15

Jupiter Rowland

jupiter_rowland@hub.netzgemeinde.eu

@𝓐𝓷𝓭𝔂𝓣𝓲𝓮𝓭𝔂𝓮 𓀤 @-0--1- @David G. Smith You mean as in someone else permanently writing an alt-text into the original image?

Good luck wrestling down the permissions system and acquiring permission to remotely alter the original image post if it is from Hubzilla, (streams) or Forte. Not everything in the Fediverse is Mastodon or works like Mastodon.

#FediMeta #FediverseMeta #CWFediMeta #CWFediverseMeta #Hubzilla #Streams #(streams) #Forte #AltText #AltTextMeta #CWAltTextMeta

2025-04-20 17:36:40

-0--1-

_9CL7T9k8cjnD_@mastodon.social

@jupiter_rowland @iDGS Not someone else. AltBot. It generates the description very well and posts it below. Then it gripes at me because I don't add AltTxt. I understand there are permission issues, but I think it would be a nice, logical step for either (1) AltBot to write the text to the AltTxt rather than post it below, OR (2) when an image is posted, and AltBot writes the description to the AltTxt. It would cut down on network traffic and memory.

2025-04-20 17:37:11

𝓐𝓷𝓭𝔂𝓣𝓲𝓮𝓭𝔂𝓮 𓀤

andytiedye@sfba.social

Of course you can't alter the original image, but if you can put the alt text into the image when you share it, then it will still be there when it is copied and shared again.

The creator might even add alt text to the image when creating it, if there were some standard way to do so.

@jupiter_rowland

2025-04-20 17:37:37

Jupiter Rowland

jupiter_rowland@hub.netzgemeinde.eu

@-0--1- By the way, I'll accept that AltBot is

AMAZINGLY GOOD

when it's better at describing and explaining images about extremely obscure niche topics accurately than experts on these topics with years of experience.

I've yet to encounter an AI that outdoes my own image-describing in accuracy and level of detail. This, by the way, is likely to require knowledge that only I have.

CC: @David G. Smith

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

2025-04-20 17:42:01

-0--1-

_9CL7T9k8cjnD_@mastodon.social

@jupiter_rowland @iDGS Good point. The MAIN reason I don't add AltTxt and let AltBot do the work is because I couldn't begin to describe some of the stuff I post. I LOVE AltBot. Sometimes, it's quite funny.

2025-04-20 17:56:28

-0--1-

_9CL7T9k8cjnD_@mastodon.social

@jupiter_rowland @iDGS "This, by the way, is likely to require knowledge that only I have." You are spot on. In order for an #AI #LLM to match your descriptive ability it would have to be TRAINED on everything you have been TRAINED on. That's one of the reasons I prompt multiple AI/LLMs and compare the responses. Their training sets and architectures are all different.

2025-04-20 17:56:49

Jupiter Rowland

jupiter_rowland@hub.netzgemeinde.eu

@-0--1- @David G. Smith Still, first of all, if I posted an image without an alt-text (which I'd never do), AltBot would have to assume full admin rights over the Hubzilla channel that I'm currently commenting from because that's the only way for another Fediverse actor to alter the source code of my posts.

Altering the source code of the post is necessary because Hubzilla, (streams) and Forte neither have a dedicated alt-text field, nor are images file attachments there. Rather, images are embedded directly into the post, in-line, just the same way blogs handle images. And alt-text has to be woven into the image-embedding code in the post. Thus, the post itself has to be altered.

So, assuming AltBot actually manages to circumvent the two most advanced permissions systems in the Fediverse, it would have to trace back an image that it perceives as a file attachment to where exactly the embedding code for that particular image is in the post.

It would have to be able to both understand and write the specific flavour of BBcode used by Hubzilla, (streams) and Forte.

It would have to, for example, take this piece of code...

[zrl=https://hub.netzgemeinde.eu/photos/jupiter_rowland/image/b1e7bf9c-07d8-45b6-90bb-f43e27199295][zmg=800x533]https://hub.netzgemeinde.eu/photo/b1e7bf9c-07d8-45b6-90bb-f43e27199295-2.jpg[/zmg][/zrl]

...and edit it into this.

[zrl=https://hub.netzgemeinde.eu/photos/jupiter_rowland/image/b1e7bf9c-07d8-45b6-90bb-f43e27199295][zmg=https://hub.netzgemeinde.eu/photo/b1e7bf9c-07d8-45b6-90bb-f43e27199295-2.jpg]Digital shaded rendering of the main building of the Universal Campus, a downloadable island location for 3-D virtual worlds based on OpenSimulator. The camera position is about three metres or ten feet above the ground. The camera is tilted slightly upward and rotated slightly to the left from the building's longitudinal axis. The futuristic building is over 200 metres long, stretching far into the distance, and its front is about 50 metres wide. Its structure is mostly textured to resemble brushed stainless steel, and almost everything in-between is grey tinted glass. The main entrance of the building in the middle of the front has two pairs of glass doors. They are surrounded by a massive complex geometrical structure, very roughly reminiscent of a vintage video game spacecraft with the front facing upward. Four huge cylindrical pillars carry the roof end, the outer two of which extend beyond it. All are tilted away from the landing area in front of the building and at the same time outward to the sides. The sides of the building are slightly tilted themselves. In the distance, a large geodesic dome rises from the building. There is a large circular area in front of the main entrance as well as several wide paths. They have light concrete textures, and they are lined with low walls with almost white concrete textures. Furthermore, various shrubs and trees decorate the scenery.[/zmg][/zrl]

Not to mention that AltBot would require extensive detail niche knowledge about the topic covered by the image to be able to whip up the above alt-text in the first place. (By the way: The alt-text example is genuine. I've actually used it. And it's an extremely whittled-down version of the long image description of the same image in the post itself, a description which has to be the longest in the entire Fediverse.)

Ideally, AltBot would do so without flagging the post as edited.

#Long #LongPost #CWLong #CWLongPost #FediMeta #FediverseMeta #CWFediMeta #CWFediverseMeta #Hubzilla #Streams #(streams) #Forte #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta

2025-04-20 18:01:22

-0--1-

_9CL7T9k8cjnD_@mastodon.social

@jupiter_rowland @iDGS " Thus, the post itself has to be altered." So, then the solution is to incorporate AltBot into the process when a post is made. That way external permissions are the express control of the original poster. Perhaps have the AltBot generate the AltTxt then give the original poster the copy to review prior to commit.

2025-04-20 21:42:30

Jupiter Rowland

jupiter_rowland@hub.netzgemeinde.eu

@-0--1- @David G. Smith If anything, the AI to describe the image should be chooseable, and the available AIs should be configurable at least for the admin. And especially, AI image description must not be mandatory and hard-coded. There must always be a way to describe an image manually, no matter how many people swear that AI is better at describing any image out there than any human.

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