Content Aware Fill Assistance

Nothing here to fill. the video is blurry as heck and does not have any discernible patterns a AI-based fill algorithm could latch on (well, until ten years down the road it has "learned" enough such scenarios). Aside from that your workflow doesn't really make much sense in the first place. Why turn something that only requires duplicating the layer, masking out suitable areas and overlaying it onto the original into a complicated exercise? Unless there is soemthing going on you are not telling us, this is something where good old basic techniques wioll work just fine. Chances are you could even create the required overlay snthetically with effects like a stretched out Fractal Noise colored suitably. Again, I think you are unnecessarily complicating this and trying to use the most unsuitable approach. Give that masked overlay thing a try...

Mylenium