hallenbeck, 29 days ago

@cheeaun @stefan I post a lot of text based screenshots. If Phanpy had the OCR thingy, I'm sure I'd add the alt text more often. I usually try to copy paste from the source, but when mobile and in a hurry I sometimes can't for various reasons (mobile UI for copy and paste being terrible is one; sites disallowing copy paste another). I'll sometimes use Google Lens too. But OCR build in would be the best.

cheeaun, 28 days ago

@hallenbeck have you tried other OCR solutions besides Google Lens? Mastodon Web actually has it (tesseract.js) but it's quite clunky and only supports English. It's possible to add more languages but file sizes are really huge to be downloaded.

Also on Mastodon, there are video and audio attachments too, so will need a separate "OCR"-equivalents 🫠

@stefan

jeromechoo, 29 days ago

@cheeaun @stefan does this count unlisted posts? I don't always alt-text unlisted post replies.

stefan, 29 days ago

@jeromechoo @cheeaun No, just public ones.

stefan, 29 days ago

@cheeaun Hmm, I wonder what impact it would have to change the placeholder text to something like:

"Imagine your dad is blind and won't be able to see this picture. How would you describe it to him?"

🙃

See also: https://github.com/mastodon/mastodon/issues/27731

cheeaun, 29 days ago

@stefan I did have some thoughts on that, though the intention would be to encourage writing proper alternate text as I found that some people treat it like captions — not describing the image but adding non-contextual text instead.

stefan, 29 days ago

@cheeaun Yes, treating alt text as image caption is a great way to put this, and something I've been wondering about myself. https://stefanbohacek.online/@stefan/112390614475468330