Okay, memory shouldn't be the problem, but 17-20 images most certainly could be. One of the reasons Gigapixel has so many more models is because it analyzes each image for differences. Add that to the new facial recovery feature in v6.1 and you could be choking the program from time to time. When I asked for the average image size, I was looking for megabytes not video formats. :) That's something else you'll need to consider. Anyway, my suggestion is to try to analyze your batches for similarity before processing. Try smaller numbers, perhaps starting with no more than 5-6 images per batch and see if your results aren't better. Good luck.