Sometimes an HTML document doesn't define its size at all, so any size
you choose is just as legitimate as any other. Look around to see if CSS
provides a hint (e.g. HTML { width: 640px; } ), otherwise choose an an
arbitrary width (perhaps 800px or 1024px would be good?), and set the height
to the closest approximation to "infinity" that you can manage, to get rid
of the scroll bar.

   Then you got to write some AI to detect where the end of the page is,
and truncate all the stuff below it. Perhaps you can add some visible
signature (an image, text, an <hr/> element) at the bottom of the HTML
document, and when the AI sees this signature in the screenshot, it knows to
cut the image there.

   - Oliver