In the meantime, it is recommended to bring in each camera clip separately to export out of Video AI. The models have a rough time transitioning between clips and cuts or rapid movement.
I do understand that proper scene detection isn’t trivial, but what if people use edited material? Do you suggest splitting the media in scenes prior to processing with Topaz?
My approach to this problem is re-rendering affected parts as separate scenes. That’s usually just a few seconds of rendering , but needs checking each scene cut after rendering and Topaz provides zero support for this process.