Food for thought—you can make your tweets accessible to blind Twitter users by including a text transcript for images. https://help.twitter.com/en/using-twitter/picture-descriptions … After you activate the setting (Settings → Accessibility → Compose image descriptions), you can use this feature. And it helps.
-
-
-
Replying to @mwichary @handcoding
I don’t feel this was particularly necessary here – both my tweet and the link contain everything that’s needed in text.
1 reply 0 retweets 0 likes -
Replying to @mwichary
Just to put this out there—if the information in the image wasn’t necessary to getting the most enjoyment out of your tweet, you wouldn’t have attached the image in the first place, eh?
1 reply 0 retweets 1 like -
Replying to @handcoding
Not necessarily. I included the image first, and then decided to paste the text inside. Kept the image more for extra visual context if you click through, and because I felt there was no harm.
1 reply 0 retweets 0 likes -
Replying to @mwichary @handcoding
Now, I’m going to throw the ball over to you. :·) Would you mind contacting Twitter and suggesting they automatically OCR text in images or provide a more forceful description prompt if they detect text? This way this can be solved en masse…
2 replies 0 retweets 0 likes -
Replying to @mwichary
For what it’s worth, modern OCR of tends to work best when the source is 150+ dpi. At the 72 dpi of the web, there’d be far too many errors to be useful. :( (If you’d like something more assertive—if you follow
@PleaseCaption, it’ll remind you if you leave off a description.)2 replies 0 retweets 0 likes -
What makes things more complicated is that attached images with text aren’t always screenshots—sometimes they’re photographs of storefronts or pics of temporary paper signs. And sometimes attached images don’t have text in them at all—like maybe if they’re kittens or doggos.
1 reply 0 retweets 0 likes -
Replying to @handcoding
I think that’s okay. Modern AI/ML can take care of a lot of that. Pragmatically, I think solving it this way would be much more impactful in the longer run than even the best UI prompting people to type things in.
1 reply 0 retweets 0 likes -
Replying to @mwichary
Modern AI/ML may be able to solve it in time. But not unlike engineering toward self-driving cars, I think that meaningful image descriptions through AI/ML may fall victim to the 90/90 rule—we may be “five years out” for the next fifteen to twenty years. https://en.wikipedia.org/wiki/Ninety-ninety_rule …
1 reply 0 retweets 1 like
Maybe. But it will *definitely* be fifteen to twenty years if no one asks. :·)
-
-
Replying to @mwichary
Accessibility advocates would love to solve this, and there will be heaps of praise to whoever gets this nailed down. So it’s not for a lack of demand. And in the meantime, I don’t mind pasting the graf from the webpage into the screenshot-description box.
0 replies 0 retweets 0 likesThanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.
