It was pointed out to me that I forgot to pass "keep=True" to blankclip. Since the script is memory bound, passing that parameter pretty much just doubles the performance, so the fps numbers quoted above are about half of what the filter itself is actually capable of.
|