I don't know what they're using, but I hope it gets better soon. Particularly with Asian character sets it's pretty hit or miss. Then put the result through Google Translate and... wow (not in a good way). Love the concept, just hope there's a 2.0 in the near future!

I was very disappointed in my results:

https://docs.google.com/document/pub?id=17-AYllk3J2SRcIzHKpf...

Does this image have a resolution or a font problem?

Any idea if google is uing using http://code.google.com/p/ocropus/ to do this?

I was under the impression that was what they used for the book scanning project.

Maybe they're using this: http://code.google.com/p/tesseract-ocr/

I wonder how long it'll take them to implement automatic language detection like what we see in Google Translate.

Is there an API for this?