Well, the „alien“ text issue is there since as long as those advanced AI models exist.
In TopazPhoto it was the same- until we got a feature to „preserve text“.
I’m still waiting for this to hopefully arrive in Topaz Video as well someday. The code is already there so it should be possible to enable such in the Video upscaling as well
This is being worked on by the research team just as it was in Photo AI. The conversion from a photo model though to one that handles video frames and tracks correctly across the frames does take some time to develop.