My super-simple app WordOff has an API for cleaning up cruft from Word-exported HTML. You could override the save method of your flatpages model to pipe your HTML through the API the first time it gets saved. Something like this:
import urllib
import urllib2
def decruft(html):
data = urllib.urlencode({'html' : html})
req = urllib2.Request('http://wordoff.org/api/clean', data)
response = urllib2.urlopen(req)
return response.read()
def save(self, **kwargs):
if not self.pk: # only de-cruft when content is first added
self.content = decruft(self.content)
super(FlatPage, self).save(**kwargs)