I have a problem. I wanted to post my art here, but I feel overwhelmed by the need to add an image description every time. Since my images are fantasy and scifi themed, describing the content in a meaningful manner often is very time-consuming and difficult. And because in the #Fediverse I get yelled at if I do not add an image description I am highly frustrated. So I have lots of images that were not posted here.

Any idea how to solve that? Insights? Ideas?

Obviously a good solution would be to add a mechanic so others can describe an image and the image poster can approve (or deny) a description that then is added. But that would have to be done at protocol level.
I expected what did happen. A FOSS guy nerdsplained to me that if I want a feature I have to contribute code or pay for it. That is not solution oriented, that is kicking the problem under the bus and this kind of dev arrogance is a major problem through all of FOSS software. I work in IT for decades and I really do not need to be lectured by those guys how the worlds works (according to their views).
@xanathon you sound like the people who get angry at service workers because they didn't talk to you nice
As I just found out, this whole discussion is unnecessary anyway. If you use a screen reader in Google and opt in for it, you can get image descriptions from multiple image content recognition models for alt-less images automatically.

"If someone using a screen reader chooses to opt in through Settings, an unlabeled image on Chrome is sent securely to a Google server running machine learning software. The technology aggregates data from multiple machine-learning models."

https://blog.google/outreach-initiatives/accessibility/get-image-descriptions/
Using AI to give people who are blind the “full picture”

Google is helping people who are blind browse the web by incorporating machine image descriptions in Chrome.

Google
Yesterday I tried the Google Vision AI API to solve this. I created an account in the Google console and wrote a NodeRED-Flow accessing it where I could provide an image URL via Node RED UI and the flow then fetched information about it from the API. Unfortunately it only delivers terms about things in the image, not full descriptions, and those results are not good.

I also tested the image caption function of Microsoft Copilot (provide an image and let it write a caption) - and the results were only good for a laugh, but were completely unusable.

ChatGPT 4 may be another solution, but I am not willing to pay 20 bucks a month for it.

#imageDescription #imageAltText #Fediverse

@xanathon Ignore the yelling people and post them anyways. You're still free to do what you desire, even if it (potentially) limits the interaction of others with your posts.

Alternatively, you can just add a very technical description like "Image of my latest art".

Or, probably not a great alternative, since you may not want to feed your art into AI, but AI with image recognition could write a description.

@darkcisum I see the reason and need for alt texts and I rather would like a solution for this. AI is completely useless for this case, as it is not able the describe fantasy and scifi based art images. I tried that and had a lot of laughs. :)
@xanathon I feel like especially for art as in pictures/drawing/painting it's okay to not have an alt text. Paintings in an art gallery usually also just have the title of the piece and have people interpret how it affects them. Maybe someone would write a guide for the gallery either interpreting or just plainly describing the piece, but there aren't people yelling at the artist for not including an extensive description with the painting.
@xanathon And don't get me wrong, I think it's great that alt texts are pushed much more on here.

@xanathon I'm not a direct user of it, but https://pixelfed.org/ might have something for that in theory and it is a tool primary for art and photos, so probably the right tool for you anyway.

Regarding accessibility, I see the point in having more people to be able to describe it. Maybe there the comments help with this approach?
But besides that, please keep in mind what burden it is to those impaired to see the images and not being able to resolve the issue on their end.

Pixelfed - Decentralized social media

Learn more about Pixelfed, the free and open-source decentralized photo sharing social media platform

Pixelfed
@mohs Thanks for the suggestion, but I had a Pixelfed account and moved away from it because of multiple bugs that were not acknowledged at first (only later after multiple reports) and that lead to images not being federated (among other problems). I went away from it and will not go back, since the devs and admins were not friendly or helpful, so I no longer have trust in that project. Also Pixelfeds UX and UI leaves a lot to be desired.
@xanathon I know one person who just puts “don’t bug me about alt text” in their profile.
@MisuseCase I see the reason and need for the alt texts. So I am looking for a way to solve this instead of just refusing it.
@MisuseCase @xanathon I would put "don't bug me about alt text" as alt text.

@xanathon

I consider it to be a suggestion, not a rule.

We might try to describe what's in it and some aspects of the aesthetic decisions made, but it's mostly not possible to successfully convey the essence of a painting with words.

It's a generous & thoughtful idea to advance (and it's not without merit to try), but the primary reason we paint is to share what can't be conveyed with words. #Painting

@xanathon what about use the same generic image description for all and every image you post?
If, for some reason, you want use a specific description for an image, you still are free to do. And no one can complain the content of your descriptions, as it is your work and if they (the yelling people) don't like it they just have to deal with it.
@dancer_xiv As I said in another comment: I understand the need for descriptions for visually impaired people and would prefer to solve this solution-focused instead of a workaround that does not help these people. That's why I made an example for a tech solution of the problem. I think a discussion for a protocol-based solution is needed instead of trying to work around it with tricks that leave visually impaired people out.
@xanathon If the problem to solve is different than the one I replied (because missing context probably), the only solution is you adding a meaningful description for each image, if you want address the visually impaired. You can't ask a machine to do that for you. Yours is a creative work, not something that can be automated. Unless you want some AI create that text for you. Nowadays people use AI for everything...
@dancer_xiv Read my opening post to this problem. I NEVER asked to let a machine do the description, as I know quite well that this is impossible. And I explained why doing it myself is stressful and frustrating. So you are quoting the problem back at me and that is not helpful.
@xanathon A possible solution is to leave it blank and ask your viewers to suggest a good description for editing it later. But don't know how edited posts will interact with the timeline of a conversation...
@xanathon Valid for any open source project: If you cannot contribute a code fix yourself, see if you can fund the development so other can code the fix or... wait. There is no epiphany here. No matter how useful something might seem to you or me, nothing is to be expected from people already graciously giving substantial amount of their free time. They often do it because they can choose what they work on based on their interests, skills & the kind of complexity they are comfortable tackling
@xanathon people sometimes reply with their own image descriptions, maybe there's something you could add to your posts to encourage people to do so
@xanathon with art especially it's kind of an open problem as to what is even useful to include as alt text. I try my best, but I think on at least one occasion I had to put "I'm at a loss of how to explain this, I am sorry" in an image description. Different people need different things, too, so I'm not sure there's a one size fits all approach anyway. Making it more collaborative somehow seems like a good idea.
@xanathon there's obviously options, like there's a bunch of AI models that can describe an image to various degrees of complexity, but honestly, just make your descriptions more concise and less wordy. You can't recreate the image as text
@PurpleBooth nope. Been there, done that, got no shirt. AI models are able to describe plain everyday pictures, but not art. That will probably improve and soon, but do I want to feed my images into AI models? Only if that runs on my computer.

@xanathon  The people yelling at you can fuck right off!

...which unfortunately won't stop them yelling at you. *sighs*

@xanathon Theoretically something like this would work already! Sometimes people add image descriptions with replies. Problem here: Happens relatively rarely compared to complaints.
Regardless if we had such a feature or use replies for suggestions, this only works as long people don't feel like you take advantage of their help. So you might have to put your own short suggestion in there as well.
@xanathon I don't know if this helps: Don't overthink it. In my own limited experience, broad descriptions without many details (unless important) are enough most of the time.
And you always can call for better descriptions, if someone has an idea, nurturing constructive interactions instead of complaints.