In my experience the motion deblur model is not usable because it introduces many artifacts, especially on hair.
Do you have any plans to develop an enhanced motion deblur model?
It would be a good complement to the stabilization model. Because of camera jittery motion, there are often few blurry frames between sharp frames. I would love a model that could replicate the details of the sharp frames in the blurry ones, instead of deblurring these frames without any reference input.
