I have used the online OCR with good results. When I use the API, I don’t get as good results. I’m using python and have used both of the code examples and gotten similar, bad results.
A support person suggested using scale=true. I tried that and got the same bad results.