So I've been very happy with MadVr's upscaling for live-action content, but I seem to have a hard time making animated content look even remotely good. My current default settings for all video are to use Jinc with AR for chroma and luma, with debanding, and occasionally NNEDI image doubling when my GPU can keep up with the content.
For whatever reason, those settings seem to make animation look like hell with tons of mosquito noise and/or ringing (I'm not fully sure how to describe it, but I see visible pixelation around object boundaries and fast movement of small objects looks juddery and terrible). The extent of the problem seems to vary quite a bit from series to series, but it seems especially bad with 1) interlaced animation, 2) on any kind of onscreen text, 3) in scenes with small moving objects.
For example, I have been unable to find settings that make "the Simpsons" look good, although the problem seems more pronounced with older seasons of the Simpsons than newer ones (I'm watching the second season right now and it looks just terrible). And I've seen the same problem in lots of animation.
To be clear, I checked the OSD and I'm getting no dropped or repeated frames, and I tried watching direct from DVD to rule out any kind of file-corruption/re-compression issues.
Question 1) Has anyone found "best practices" for upscaling animated content? I've been toying with experimenting with softer upscaling algorithms, but I'm not sure if that's the issue, or if it's something else?
Question 2) I've had good luck with MadVR profile switching based on content type (i.e. interlaced or not), but that was all based on info MadVr would already have; I'm not sure what the best way to signal animated content to MadVR would be. Assuming I can set up an alternative MadVR profile for animation, is it possible to get MadVR to switch profiles based on tag information, or could I include a string in the filename to get MadVR to switch, or something else?
Any advice is appreciated.