• Apps
  • How to watch AI Passthrough vids with HereSphere or locally?

Thanks for that, Koreanman, that was on my list as something to try. Could that be done with the free version of DaVinci? Just by watching the SLR AI passthroughs, I'm pretty sure the matte streams are lower resolution than the source material, so a 1K or 2K matte video might be acceptable on an 8K video. I think DaVinci has a free-version 4K export limit. I don't think I'm going to render hours and hours of video, but I was thinking about creating PMVs and doing select clips.

    sunkenlust I understand that DaVinci Resolve's Magic Mask requires the Studio version to be installed. The Studio version is quite expensive unless you're doing professional video production work. Therefore, I recommend searching the internet for ways to try it for free before making a purchase.

      Will be great if someone make working video background remover Addon for Automatic1111 soon ... I have been tried some but it seems it doesnt work proper with 8K vids so far ...

        boboweb why would you want to put it in there? Doesn't seem like the right tool to me.

          fenderwq

          Because its free and usually fast ... But i dont mind if there is some other dedicated standalone github app ...

            boboweb Just saying you'd probably be better off just to do it in the command line (or a bat file) and process everything in small sections. Creating masks takes a long time and it is good to be able to resume where you left of or stop processing if you need your pc for something else etc. You can also just process all videos in a folder using this way stuff like that. No real advantage of putting that in Automatic1111 since there is no real use interaction/ different use case.

            fenderwq I think DeoVR does, that is how SLR can even do AI Passthrough, Alphapacking, Chromakey.

            a month later

            If someone care to try RVM i get the best results with these parameters:

            1.)using rvm_resnet50.pth model instead the dafault rvm_mobilenetv3.pth - i get too many transparent spots with mobilenetv3 in composition output
            2.)set downsample_ratio 0.1 if its higher it get more transparent spots and its slower
            3.)seq_chunk 4 using about 12GB VRAM with 8K vid , slightly faster then seq_chunk 1 - about 1.9s/it

            in summary i can say RVM works good but unfortunatelly its slow . It takes about 150 mins for 90 seconds 8K vid on my 4060Ti 16GB my CPU is very old indeed . I guess the models are too old and are not good optimized for todays hardware and HQ VR vids .