OCR.Space wysiwyg landing page demo gives far better results than my python implementation!

Hello everyone,
I’m really new to using OCR.Space. First I would like to thank you for releasing such a great product!
However, I’m facing an issue. I first tried the OCR using the “Try it” form on OCR.Space’s landing page (https://ocr.space/#) and got a perfect result reading a french administrative document (standardized and clear font).
I then when to try the api with a python script. I got a result, but it is not good at all, there are mistakes at every word :confused:
I used Zaargh’s github repo (as advised by OCR.Space) and copied the functions he wrote. I took care to put my API key and changed the language to french (language=fre). The parsed text obtained with language=fre is even worse than with the default (language=eng). Could someone help me please?
How could I obtain the same perfect result I got when I tried the ocr in the landing page interface?
Thank you in advance for your answers, feel free to ask if you need a copy of my code.
Have a nice day
Ugo