did you know that web UI has OCR built in? so when you are posting screenshots of text, a transcription of the text to put into a description of the image is literally one click away. when uploading image into a toot, just klik 'Edit' and then 'detect text from picture'. that's it.

please help sighted people. it's costing you two extra clicks.

@luka hmm I do describe many of the images I post but I never noticed this 😅
Doesn't work on the instance here… maybe it misses some package (tesseract?).

@mmu_man @luka

I’ve also been having issues with this on for months, so it might not be your instance’s fault:

cc @Gargron

@luka except for some reason on cathode church it just fails with a "RuntimeError: abort(undefined)". I haven't been able to work out what's wrong.

@luka oh neat, last time I’d looked the thread was still unresolved. I’ll have to bring up to main on glitch-soc again soon

@luka Why not using the already existing IPTC/EXIF instead? At least the description, titel and photog tag among those meta data. Saves a lot of energy btw. ... 🖖

@luka Almost every photographer tags and describes his/hers photos in CaptureOne, Lightroom ... with $photographer, $description and $title. These data are in the photo in the EXIF/IPTC.

FB, Insta ... strip this data - incl. copyright - to obfuscate the origin due to business reasons.

Unfortunately, the Fediverse has joined this folklore.

It is also not fair to visually impaired people. (

Furthermore any processing (again) on external machines cost energy/CO2 ...

@luka Not to forget, that the image description holds an extra 1500 letters as opposed to 500 in the standard Mastodon toot. Perfekt for hidden messages, too.

Sign in to participate in the conversation

SoNoMu (Sound Noise Music) is a mastodon instance for musicians, sound-artists, producers of any kind of aural noise, songwriters, bedroom producers, sonic manglers and algorave livecoders. -> more...