Need I say more?! 
The gigapixel version is northing short of AMAZING and it would be great if we can apply the same REDEFINE / RECOVER generative models and introduce some consistency, as slow as it might be.
Same for generative remove.
Possible?
This would be a great feature, but I suspect that instead of measuring performance by frames per second it would be seconds or minutes per frame.
Probably right, still might be great!
Another idea would be to do it every X number of frames and apply a similar approach to EbSynth between the upscaled clip to the redefined clip to fill in the detail
With todays consumer hardware it‘d surely be minutes per frame, but:
We had such to some extent with the earlier 2.6 versions of Video Enhance and could cope with that.
Plus, even given a processing time of 1-2 mins per frame on high end hardware, it still could be feasible for smaller clips.
