New model type for interlaced video?

I’m on a posting spree and wanted to mention something that maybe the devs could look into adding. I upscale alot of interlaced DVDs and sometimes DDV looks great, sometimes not so much…it goes crazy on dark things like eyebrows and does not look good. All the Dione models will do this on certain videos for whatever reason and the dehalo models are usually too strong.

I found a trick for making these look right. I use Linux for anything other than my veai systems and one simple but powerful video editor is Avidemux. Avidemux has a filter called “stack fields” where it takes the even fields and stacks them together, then stacks the odd fields together under it. So it basically looks like a squished mirror image top and bottom. Gaia brings out alot of detail without making stuff too bold/sharp. If I run the stacked video with Gaia then stuff it back into Avidemux and run “unstack fields” it produces a much better looking interlaced video which I also have to run an Avidemux de-interlacing filter on. But then it’s only 30fps and I have to run it through flowframes.

It’s a huge PIA but sometimes when Dione looks like trash this looks waaay better. I’m wondering if there is some possible way to add a VEAI model that splits the fields and stacks them to run something like Gaia, then puts them back together and runs a frame doubling deinterlacer model that does not go crazy on the sharpening. Deinterlacing the video into progressive beforehand does not produce the same results. I typically go after interlaced DVDs for stuff that has not been updated to Blu-ray because one way or another VEAI can typically make them look like they were originally 1080P videos. Having to do the above is just a rail PAI that I’m sure loses something by being re-compressed 4 times.