Why not include all models in the list for interlaced sources? All you’re are doing is using ffmpeg’s inbuilt deinterlacing for Iris and Proteus etc. (e.g. adding “bwdif=mode=0:parity=-1:deint=0,” at the start of filter_complex) so why on earth omit Artemis and others from the list for interlaced sources?
Users shouldn’t be expected to resort to the CLI and so they’re being restricted in their choices by a totally illogical design decision.
Thanks.
Andy
Deinterlace & Inverse Telecine (and ideally idet) should be a separate “enhancement” panel in the UI. Even if it is just a “preprocess” panel that contains the useful non-neural network pre-process filters (like dedot, which is very useful for removing dot-crawl and rainbowing).
It is all very strange that there are marketed as “deinterlacing” models, but really these models are ones that “have been optimized for deinterlaced sources”. But calling them “deinterlacing” models is really just confusing. bwdif is doing the deinterlacing - and it would be much clearer if this was broken out as a separate pre-processing function.
I usually deinterlace first in FFmpeg and save as FFV1 lossless (rgb48le), at the cost of temporary storage but for the benefit of not deinterlacing and converting colorspace to rgb48le.
I tend to use the TVAI CLI for my final process anyway. I wish one could edit/tweak the final command-line in the app, without having to edit the command in a text editor and then reverting to the terminal.
