I have a 13:23-long .mov that I converted from 1920x1080 to 3840x2160 in Topaz Video AI 5.3.6. I then took the output of that process and fed it back into Video AI for stabilization (rolling shutter and jittery motion both checked, as well as motion deblur). The original upscale took about 8 hours. This stabilization is taking about 16 hours.
My system: tower with AMD Ryzen 9 5950X 16-Core Processor @ 3.60 GHz; 128 GB RAM; Win10 Pro 64; NVIDIA GeForce RTX 3070 Ti; multiple 2 TB NVME banks.
Sort of just the nature of these things. Certain things (effects) just require more computation power, if you want it done faster you’ll just need something more powerful. From what I understand those Nvidia workstation cards like the Rtx 6000 and the like are insanely good at encoding and decoding, but they’re specialized cards for those type of tasks. It’s a shame that even some of the lowest tier ones basically cost an arm and a leg.
Running the stabilization and the motion deblur model will take awhile, than adding on the rolling shutter and jittery motion adds more time on. Depending on how many passes you put for the jittery motion it can compound even further.
They are very heavy tasks that require each frame to be analyzed and processed, so even with a built-in system, they will take time.
This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.
