Online test result varies from the API tests


Hi support,

I just tested out the enclosed 2 PNGs. The online and API gave me the relatively good and far less good OCR results respespctively. Would you figure out for me why?


P.S.: Unfortunately I am a new user so being limited to post only one image.


Moreover, the API OCR result is also inconsistent from image to image. For example, ‘Phone / Fax’ is recognizable from some images while not recognizable from the other images.

To hear from you!

When I am trying to convert PDF file online - the results are much better then when I am using free API OCR. Is it because it is free api and not PRO or all APIs should return identical results?

The likely reason for the difference between online ocr form result from on the front page and your API call is that the online form uses the parameter scale=true in its API call.

So if you use the OCR API, you need to set the scale parameter to true in the call. When the parameter is missing, the default is scale=false. Once you do that, the results should be identical.

scale [Optional] true/false

If set to true, the api does some internal upscaling. This can improve the OCR result significantly, especially for low-resolution PDF scans. In the front page demo, this parameter is set to true.

Nevermind, I solved the problem with further online researches! I atually found the answer to my question on this forum: API OCR quality is not same
I just added scale=True to the arguments of the ocr_space_file function.
def ocr_space_file(filename, overlay=False, api_key=‘your_api_key’, language=‘fre’, scale=True):
etc …