Gave this a try earlier - had to install their demo Python to get a "real" text file output (ie with words assembled into paragraphs); it was both slower and less correct than `tesseract` for my sample files (an alumni letter, a shopping receipt).

https://en.wikipedia.org/wiki/Tesseract_(software)

It's a powerful service, but the requirement to let AWS keep your documents is a powerful disincentive for many companies, no matter what the rest of the privacy policy says. The Azure text recognition container can be hosted locally and in that case doesn't send document contents to Microsoft.

Which service are you referring to?

I found this:

https://docs.microsoft.com/en-us/azure/cognitive-services/co...

And this:

https://azure.microsoft.com/en-us/services/cognitive-service...

But a quick scan doesn't reveal local installation options.

N.B., the service limits are quite low for many types of SaaS applications and customer support has no ETA on increasing them: https://docs.aws.amazon.com/general/latest/gr/aws_service_li...

The limits do seem low (one doc every 4 seconds for synchronous calls). However, from the doc you linked:

"Amazon Textract has the following limits that you can change."

So I think you can just file a support ticket and get these limits changed.

Here's limits that can't be changed:

https://docs.aws.amazon.com/textract/latest/dg/limits.html

Only Northern Virginia, Ohio, Oregon, and Ireland?