Hacker News new | past | comments | ask | show | jobs | submit login

Now if there were some universal standard for embedding alt-texts into a picture that would also survive right-click "copy image" and paste-uploading... not just for social media (where you'll get blasted for forgetting one even just one time) but also for websites and newspapers where the image description has to be edited manually in a CMS and is, again, not attached to the picture itself...



There is. The problem is that automatically attached alt-text isn't particularly useful.

You don't know what the intention is of an image. Is it of a banana? A specific banana? The sticker on the banana?

https://shkspr.mobi/blog/2023/07/should-you-embed-alt-text-i...

And, no, this isn't something which AI can help with.

https://tink.uk/thoughts-on-screen-readers-and-image-recogni...


Fantastic articles, thank you.

I work with alt text a lot. I used a hinted LLM prompt for the first pass, and then I edit if needs be. Sometimes it misses obvious details, other times it sees things (most) human eyes would never see. Like all tools, just have to be careful how you use it.


My use case is "someone shares a dank meme on twitter, i save it to share it on bsky", and every time I have to copy the alt-text myself.




Join us for AI Startup School this June 16-17 in San Francisco!

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: