The new generative models for Gigapixel are quite impressive. While it would likely take a good deal of processing time, there isn’t, as far as I’m aware, any video/temporally stable generative model way to enhance videos. The results of single frames processed via the “recover” model in Gigapixel look MUCH better than any frame processed by even the latest Rhea model in TVAI.
very good idea
sometimes we need something like this
1 Like
From my experience a mere imitation of what you what can slowly, very slowly be achieved by dumping all the frames using Topaz Video Ai’s image frame (Encoder Image Format) export option then batch processing them all with (in my case) Topaz Photo AI and then reassembling with a video editor like DaVinci Resolve since Topaz Video AI can’t do this (or maybe it can, I have an old 3.0.12 version.)
I took weeks for 2h 4k video, I’ve tested this already 
But I’m sure there is as lot of potential for optimization to reduce time; only the changes from one the other image have to be rendered.
