INTERACT FORUM
More => Old Versions => JRiver Media Center 25 for Windows => Topic started by: mattkhan on October 24, 2019, 06:23:39 am
-
when using ROHQ, is it possible to configure deinterlacing in MC directly or do I have to create a custom config so I can get access to the LAV settings?
(context is https://forum.doom9.org/showthread.php?p=1888544#post1888544 in which I have an deinterlacing issue which only seems to happen when madvr deinterlaces, I can't recreate in mpc-be but that seems to be because deinterlacing is done by LAV where MC seems to push this onto madvr)
-
LAV settings are not accessible in MC by default ... it is preconfigured for stability reasons
Only deinterlacing option to my knowledge is in MadVR
If you could describe here what problem you are having ... (seiving through forum you link is a bit difficult as there are so many other problem posted on MadVR general forum)
-
doesn't that link take you directly to the post with the screenshots showing the problem?
It is basically deinterlacing behaving in a not obviously deterministic fashion (i.e. not driven by content)
switching to a custom config so I can configure LAV myself does resolve the problem. I'm not sure why hiding deinterlacing config contributes to stability though, doesn't seem any different to choosing whether to use hardware decoding (which is exposed).
-
There are two things you can do...
First, madvr (so you have to use ROHQ) deinterlacing can be forced by modifying the file name. I have found that adding one of the following strings to the filename will fix problems, depending on the content:
deint=film
deint=off
So for example, a file named "movie.mkv" would be renamed "movie deint=film.mkv"
One movie might require film, another might require off. There are many ways the movie can be encoded so you will have to experiment.
Second, your graphics card deint settings have a big effect on madvr. For AMD cards, I have found it is best to disable automatic deinterlacing and force the deint mode to vector adaptive. Not sure about the Nvidia equiv, but check your settings.
Try these and see if you get a good result.
-Will
-
doesn't that link take you directly to the post with the screenshots showing the problem?
It is basically deinterlacing behaving in a not obviously deterministic fashion (i.e. not driven by content)
switching to a custom config so I can configure LAV myself does resolve the problem. I'm not sure why hiding deinterlacing config contributes to stability though, doesn't seem any different to choosing whether to use hardware decoding (which is exposed).
from screenshots looks like you using dxva ... turn that off ... then try film mode
Like wer ... I try different different deinterlacing settings to determine what works best for particular content ... then move file to different folders (/film mode ... /video mode) then apply setting in madVR base on location
like madVR forum mentioned ... deintetkacing technique depends on the content ... whether it was shot interlaced ... or was shot progressively then converted to interlaced ... later process can be reversed with IVTC (film mode)
usually IVTC should give perfect result ... but often content is edited after it was interlaced ... so it’s not so simple and results will never be perfect (few glitches there and there)
I have PAL futurama that theoretically should give perfect result with IVTC ... albeit it was mastered so badly that madVR struggle to give good result (in some scene MadVR did great ... in others LG C6 did better ... drove me nuts) ... then I hunted down US NTSC version and madVR deinterlacing look great (not perfect ... but hardly noticeable from normal viewing distance)
PS. also screenshot looks weird ... frame rate is 25 which suggest film mode is being used but there is no cadence detection ... if video mode is used frame rate for pal should be 50 ... must be dxva thingy ... try turning dxva off
-
the explanation for using 25fps is further down in the linked thread (HDBaseT limitation until I can run a new HDMI cable)
I appreciate the suggestions but, as I said in the last post, I have already solved the problem by changing the LAV config.
from screenshots looks like you using dxva ... turn that off ... then try film mode
how do you turn DXVA off? As I understand it madvr will either use the GPU (which means DXVA doesn't it?) for video mode or its own IVTC implementation for film so it's not obvious to me how you would actually turn it off (apart from forcing film mode).
-
Video mode is using dxva as I understand it ... does forcing film mode makes it better? ... OSD will show cadence detection ... if it can reliably detect cadence (like 3:2) the result will be perfect
You can also use custom setting in video option ... then specify LAV filter (separately installed from MC) and MadVR (separately installed from MC) and your choice for subtitle renderer ... in that scenario you can edit separately installed LAV settings ... you won’t get support from developers if things don’t work in this mode though