Want to OCR an image of a Dutch textbook, but chatgpt is being very inconsistent. The images are of good quality. It didn’t work on the macos app (somehow it cannot use tesseract?) but it worked on the browser version and gave perfect output. For 2 images. Now, its not working at all anymore. Its giving the error that theres a missing language package on the backend server. But why did it work before… Anyone having similar issues?
1 Like
Yes, Tesseract OCR does not have all the languages installed in the Python environment. Also there are better libraries out there like Kraken, for images of handwritten notes.
This message says to me:
(“There was a problem trying to process your image with the OCR tool, as the necessary Spanish language file was not found to convert the text. I can still help you by reviewing the image manually and transcribing the content if it is legible. Give me a moment to review the text and transcribe it.”)
Then it says it extracted it manually. But it confuses my handwriting and has many mistakes.
I’ve tried several times and nothing.
So maybe with other languages, it is not as advanced. I had never tried this tool, I was thinking that I would finally understand the doctor’s handwriting, but if it can’t understand mine, even less the doctor’s.
Hey @taniajorgeytico
Did you try forcing it to use its vision capabilities instead of writing python code and executing it?
If so, how did it go?
YES. I’m trying to access via API and it just never gets it (Hebrew). I’m pretty sure it used to work way better but I feel kinda gaslit here (bc why would performance suddenly drop…)
Huh, weird. I actually use it through the API as well and I am getting perfect results for handwritten text too.
Are you making sure that you’re uploading high quality and resolution images and setting the fidelity image understanding parameter to high?
@snww and @n001spartan
Can you send images that I can use to try to reproduce the same behavior on my end?