Sargoth

Hey, fedi knowers, quick question:
when you click the "read from image" button in the alt text dialogue box on the main branch, what happens? is it an old style OCR, or should I be worried a daemon is accidentally summoned?

Since this a quick question, a quick answer will suffice. Since this is fedi, I kinda can't stop anyone from infodumping, but I neverdeleuze hereby release you from the obligation

Mar 27 at 11:57pmWeb
Show threadaschmitz1d ago
@sargoth It loads a copy of Tesseract.js (https://tesseract.projectnaptha.com/) and does OCR locally. Tesseract 4 (which I assume that uses?) has an LSTM neural net, which is somewhat newer than some of the old pattern recognition stuff, but still fundamentally decade-or-so-old tech, not an LLM / VLM / whatever, if that's the concern. Of course, it does still summon some extra WASM code to your computer, so how you feel about the whole thing is up to you.
Tesseract.js | Pure Javascript OCR for 100 Languages!

Show threadaschmitz1d ago
@sargoth Despite the information in that toot, rest assured that you were spared a significant amount of infodumping.
Show threadSargoth1d ago

@aschmitz aooo, old style neural, which at no point has caused a CEO to go "we need to pivot immediately"

this is reassuring. thank you :3