@Nancy50 might care to comment but I think the caption is, in fact, “image”. Not very helpful. It takes a real, modern, screen reader to say; traditionally the information was in the “alt-text” associated with the “img” in an “href” tag. It’s probably still referred to as “alt-text” in techiespeak but I gave up trying to get people to care more than 25 years ago.
@ClaudnDaye - I’m not sure what you mean by “inside”. Can you post an example?
Eh… Sorry; people who have no visual function can’t see images. They can hear.
To alleviate this problem we put textual information inside image files so that screen readers might hope to be able to extract information and present it so a person with no visual capability can hear a description of what the image is. This doesn’t help much, if at all.
When HTML was created it had an additional facility, called “alt text” where the information from the image, or information added with the image, could be displayed (audibly) to a person with no visual capability.
Both your and Eric’s images have been completely stripped of all text; there was certainly text there when you and Eric generated them, but it’s been stripped. There is no standard for “alt text” within an image file. Consequently everything I did 25 years ago in this area was aimed at getting the “alt text” in there. (Real aim; secondary aim getting image author copyright preserved, that was probably more marketable but neither got anywhere.)
Maybe it’s time I started using a screen reader. My eyesight is getting worse and worse; you posted a picture of a pacific tree frog, damn things wake me up every morning in spring, but I love them. All I could read on your images is, “frog”. About as informative as “image”.
Testing a bit
Here are the steps in Discourse:
Upload the image into your Discourse post.
Discourse will insert image markup that looks something like this:
](https://canada1.discourse-cdn.com/flex030/uploads/fudiabetes/original/2X/a/a31d7585ecb1d79d7e44687f00948ddbff7c6d5e.jpeg)
image.So change this:
to this:
] alone, especially the upload link:(upload://abc123def456.png)
The key part is this:
So for your frog image, your alt text is the part before the vertical bar |.
Great; it can be done. So now how do we, well, the @supreme-masters-of-fud, ensure everyone does that whenever they upload an image?
Since Discourse already supports this using the image markup — the text inside the brackets becomes the alt text:
The real question is enforcement (which would be very hard to do, tbh). Ideally, we could add a theme component/plugin or use “Discourse AI” to warn or prompt users when an uploaded image still has generic alt text like “image.” Until then, if it’s something the community wants, it probably needs to be part of posting guidance: when you upload an image, replace image with a useful description before posting.
AI can help draft the description, but the poster should still review it because good alt text depends on context.
Discourse offers a plugin called Discourse AI which Admins could add to our suite of tools.
Discourse AI could in theory help by automatically generating a suggested description for uploaded images, which users can then review/edit before posting. That helps make alt text easier and more consistent, but it probably should not be treated as perfect or fully hands-off. I don’t think there exists a completely automated, hands off solution.
I had problems following your instructions. I uploaded a BG screenshot and I got this (carriage returns inserted for readability):
I.e. the markdown does not contain the “image” thingy. Apparently I need to replace the thing before the “|” character. I did so, I used the older inglish, jbowleres, because I know computers have problems with the apostrophe, “jbowler’s”:
So can someone confirm that a screen read does show the text for the image? Yeh, it’s not very helpful text even then but, in fact, I just posted a random screenshot…
EDIT: I just checked, it is possible to correct the default “image” text. I just corrected the text on the above image in this edit. I deliberately inserted the ’ used in the modern English possessive to see if that might cause a problem…
When i open the picture, i see markup text that it’s jbowlers bg.
I think it’s probably generating the “correct” HTML. When I look at the page source in Chrome I see that the HTML of the actual page appears to be a string passed to Java (a programming language implemented by the browser) and it does seem to contain the thing before the vertical bar.
I’m not sure what clicking on the image (what you called “opening”) does; screen readers cannot, or at least should not, do that because it’s a UI action. It’s possible they do but is strikes me as a minefield.
I don’t see how we can get people to edit “image” and, indeed, the results are not good; asking visually enabled readers to get into the mindset of someone who is not is challenging.
I’m in favour of your first suggestion; use an AI to generate a summary and then have the user, who will see the markdown and maybe even notice it, correct it. Correction is required of course; your example of ChatGPT generating “frog” for what is obviously a pacific tree frog.
I don’t know how to do this, what does ChatGPT generate for my own BG image?
Anything you want it to generate. But as you observed its only caption on top of the image. No alt text.
The points that @jbowler makes are important. It raises making meaningful caption information for images technically obscure for sighted members to add for the benefit of sight impaired members.
Sorry I’m not more aware of this issue; my macular degeneration diagnosis remind me that we need to address these issues. My first thought, however, is what has other forums, especially those who target sight impaired people?
Hee hee. 
As someone who is not computer literate can it be done so that it can be easily accomplished by someone like me? It may be the answer is don’t post pics. That is fine if that’s the case.
Might it not be better to ask those who are sight challenged what would work for them first or did I miss that step?
Just wondering…
Cute Frog!!!
Yes, although not by the text-editting approach above; it’s very non-obvious that the word “image” is, in fact, the alt-text. “Insert alt text here” would be far better and, “Insert detailed description here.” would be a non-nerdy version.
The problem is generating useful text. When this was part of what I did (a very small part unfortunately) I worked out a quick sanity check; given an image and some alt-text if I read the alt-text first then I should be able to see the salient parts of the image without ever looking at it.
Some things are not obvious. For example the frog could be described as “a green frog with a yellow orange underbelly and an orange strip under orange frog eyes.” I could imagine an image based on that, but I see colour… Not everyone has that advantage. My caption, “Pacific tree frog.” is worse in a different way; I know what it is but most people seeing that image will not, they might see, “Cute green frog.” I added information that is not available on sight for most people.
In fact coming up with a good description is a whole lot more difficult than simply plonking a BG screenshot into a post. My own captions to my BG screenshot (I edited the post so there are two) cannot be called meaningful alt-text, or, indeed, meaningful descriptions.
Can AI do better? I don’t know; I just don’t use it despite having been involved in image-related computer stuff my whole working life. “Image to text” is certainly available, although when I searched Google for their implementation they gave responses which indicate they just do text extract; Optical Character Recognition (OCR) where the text has to appear in the image. There’s a lot more money in OCR than in producing even bad descriptions of images.
On the other handle the Google AI did tell me about itself:
3. Google Chrome (Built-in Image Descriptions)
Chrome features an accessibility tool that uses Google AI to automatically generate descriptions for unlabeled images on the web (useful for alt-text or screen readers). [1]
The links in that quote do work (and are harmless, well, apart from the fact that Google will track you.) I deleted the ad-link; clicking it results in an invalid URI. I also fixed the link numbers:
Bear in mind that AI downloads the images you ask it about.
Thank you so very much for your detailed response.
I believe I understand the posts now and why it is such a challenge.
I knew I was missing the whole idea and since it was about inclusion felt it was important to understand.
Thank you so much for taking the time to write such a lengthy explanation.
