The Kenburns effect is complex, so a good understanding of how it works (software-wise) will help you to optimize the results.
Explain how it works here: <explain>
So, for optimal results:
- Do not let the edges of the image show. This will always create a jerky look since the resulting image needs to be placed on top of a background image, and the best resolution is +/- one pixel.
- If your image fills the whole screen during the Kenburns effect, but still looks jerky, enable the high-quality mode using -H. This will enable sub-pixel sampling and should make the zoom or pan look smoother for very slow effects. It takes more time to run, of course! Test with the low quality switch -L to make sure everything works before you run the final one!
- Input images should be "reasonable" in size. I'm working on an auto-sizer to optimize the speed/quality trade-off. If you use default 6MB input files, it will take some time to resize each one within the kenburns loop! Smaller input images are faster, but if you're zooming in quite a bit, you will surely notice the resolution go down.