I use Topaz on the command line and use different models depending on how noisy or shaky the video is.
It would be amazing to have an API that I can use that characterizes the overall noise in the video so I could decide what to use programmatically. Just using auto on nyx doesn’t really work for me. I switch between nyx and proteus depending on noise level.
Same goes for shakiness / motion blur, but less important than noise.