View Single Post
Old 28th April 2014, 18:16   #26283  |  Link
madshi
Registered Developer
 
Join Date: Sep 2006
Posts: 9,140
Quote:
Originally Posted by mandarinka View Post
Hmm, maybe that's not a bad idea for a feature request, to have a switch that would tell madvr to avoid scaling factors very close to 1.0x (1916 -> 1920 etc...) and just pad with black bars instead.
Something like this is my short term to do list, already, together with some related changes/improvements.

Quote:
Originally Posted by truexfan81 View Post
i just did a test, i set it back to 60hz all is fine, i set madvr to change it to 1080p59 it has the same problem. In previous madvr versions this used to work, not really sure when that was broken, but the reason for trying to use 1080p59 is to remove judder without the blur caused by smooth motion.
Could you please double check by going back to older madVR versions? It would be quite surprising to me if this was really a new issue. I think you'll find that the same problem will also occur with older madVR versions, because I think the issue's probably got nothing to do with madVR, but probably your refresh rate is simply too low. But I'm just guessing here. Maybe I'm wrong...

Quote:
Originally Posted by StinDaWg View Post
I'm not sure I understand your take here. The whole reason to use decimation on 720p59.94 videos is to play it back at it's original frame rate of 23.976. [...]

I don't see why you would want to use IVTC on true 59/60 fps content like sports, that doesn't even make any sense. It's supposed to be played back at that frame rate because it was filmed that way.
Take a step back and look at the bigger picture, in this case usability: Ideally you want to just playback some content without having to worry about which cadence it has. It should be madVR's job, only, to worry about such things. Why would you have to force madVR into film mode? Why does madVR not detect such things itself and switch everything around accordingly? The reason for that is that I've simply not implemented film vs video detection for interlaced content yet. But at some point I plan to.

Now let's look at 720p50/60 content: For this I don't have to detect film vs. video, I just have to detect the underlying cadence and that already tells me everything I need to do. So decimating 720p50/60 content is much easier than handling unknown interlaced content.

Basically, madVR could already be able *right now* to automatically handle any 720p50/60 content and decimate whatever needs to be decimated. So you could enable madVR decimation for progressive content and you'd never ever have to disable it again. E.g. consider watching TV broadcasts in real time: There'll be a movie (3:2 frame cadence), then there'll be some ads (maybe 3:2, or 2:2, or 1:1 frame cadence). madVR's cadence detection should auto detect all that and can then throw away just the duplicate frames. The nice thing is that everything could be done automatically.

However, here comes the catch: There's a chance that the cadence could change back and forth multiple times in a short time period, e.g. from one advertisement spot to the next. We do not really want madVR to switch refresh rates all the time when that happens, do we? So running at 60Hz makes some sense because it can handle any cadence, and smooth motion takes take of 3:2 content, too.

Of course I understand that if you know for a fact that your content is really film only, and every part of it is 3:2 cadence, then there's no need to use 60Hz and you can simply use 23Hz instead. But this once again requires your input: You need to tell madVR that you know that the content is straight 23Hz all the way through. Maybe I should add an option for that. But for the majority of users, letting madVR do everything automatically is preferable, and that's why currently I'm staying at 60Hz.

Anyway, that's just the current situation. I've some ideas for future builds. At the moment I've just implemented 720p50/60 decimation for the first time, so please just test whether it generally works, report bugs/problems you find, and please live for now with whatever shortcomings there might be. Once we've ironed out whatever bugs there might be, maybe I'll polish the whole thing to make everyone happy.

(I don't want any suggestions/ideas at this point, just bug reports or confirmations that everything works as expected within the limits of the current implementation.)

Quote:
Originally Posted by turbojet View Post
Here's a clip that glitches while seeking and from the start: http://www.mediafire.com/download/5f...int%3Dfilm.mpg it's reproducable most of the time, just jump around. If SM is off it doesn't happen at 60hz. At 48 and 50hz the glitches happen at 72 and 75 the glitches don't happen.
Is there already a bug entry in the tracker for this? If not, could you please create one? I don't have time for madVR at the moment. So this problem will get lost and forgotten if you don't add it to the tracker.

Quote:
Originally Posted by renethx View Post
This anime clip (One Piece #444) is 1440x1080i59, obviously 2:2 cadence. When I press Ctrl+Shift+Alt+T twice (or add the tag "deint=ivtc"), the desktop refresh rate changes to 23Hz, resulting in lots of dropped frames. (The list of display modes is "1080p60, 1080p59, 1080p50, 1080p30, 1080p29, 1080p25, 1080p24, 1080p23", all of them are supported in my display.) So I have to add both tags "deint=ivtc" and "refreshRate=29".
Euwwh, why don't you use mediafire? I refuse to login to be able to download something.

Anyway, currently madVR's forced film mode is somewhat stupid in that it doesn't switch refresh rates based on the detected cadence. Implementing refresh rate switching based on detected cadence would be possible, but difficult, because the cadence can change all the time and we don't want the refresh rate to change all the time. I have some ideas on how to maybe solve this in the future, but for now forcing film mode on always naively switches to 23Hz. That's a known limitation of the current forced film mode implementation.

Quote:
Originally Posted by nevcairiel View Post
AFAIK its not limited to decimating to 24p, it can handle various patterns - it previously just encountered a few issues if it has to decimate way too many frames (ie. to 24p from 60p).
FWIW, decimation is not as flexible as it could be. In theory I could throw any duplicate fields away, but I'm a bit afraid of that. E.g. if I detect 5:5 I just decimate that to 4:4. That works fine for 23Hz playback. If I decimated 5:5 to 2:2 it could introduce severe motion stuttering if the cadence detection happens to be incorrect. It's possible (although it shouldn't happen) that a 3:2 cadence is misdetected as 5:5. Because of that I'm playing safe and just remove the fields I have to remove to get down to 23Hz. This is also the reason why decimation for progressive content (60p -> 23p) didn't work correctly. I've now implemented support just for decimating 6:4 -> 3:2 with 60p sources, and 4:4 -> 2:2 with 50p sources. Anything other than that still doesn't decimate all duplicate fields for now.

Quote:
Originally Posted by NicolasRobidoux View Post
(I am sure you are busy with other things, so I don't necessarily expect anything to happen as a result of what follows.)

The current madVR Jinc4 (which I believe has a "deblur" set more or less like madVR's Jinc3) is a rather ordinary scheme compared to Jinc3.

The exceptional 4-lobe Jinc scheme is the one I've been calling EWA LanczosSharpest, and last showed off in the context of pixel art resampling. It is only different from what you do already in the "deblur" that defines it, which is smaller and consequently makes it use a smaller disc. In other words, the better Jinc4 is actually a cheaper scheme than what I understand you are using now.

In ImageMagick notation, it is defined by

-define filter:blur=0.88451002338585141

which I would guess could be accomplished a minor modification of madVR, and would provide a Jinc4 different from Jinc3 in a worthwhile way.
Thanks, I'll have a look at this when I find some time. At the moment I'm busy with my commercial projects.

Quote:
Originally Posted by Anime Viewer View Post
seidweise over on the KCP Codec forum put out a program that detects OpenCL on systems. Looks like AMD and Intel have the lead over Nvidia when it comes to the OpenCL version implemented on their devices.
That's no new information. Every developer knows this. Nothing to see here, move on.

Quote:
Originally Posted by johnniedoo View Post
I have just noticed that I am getting significant 'presentation glitches after I switch from 'forced film' 24hz refresh on monitor and smooth motion on
If i disable auto detect , I get multiple dropped frames in the forced movie (or video) option and , of course, it forces the monitor to be in 24hz for the old movies i tend to watch.
i check all the movies/videos i watch with 'media info' f for whatever info i can get on whether it is interlaced or not and never really sure on the fps listed
when i watch these old movies -usually either isos i made from dvd mpg-2 some are interleaved, some are listed as progressive and all of that is still pretty mysterious to me, but play back and the de interlace option may become necessary . i often uncheck the madvr deinterlace all together or use the disable if not sure
I use ffdshow external filters as they have worked best for m y overall movie/video preferences and use the latest available from source forge and de interlace there when needed.
If i watch this type of movie in forced movie i get all kinds of frame drops, if i keep the auto detect enabled, and put the refresh rate to 60hz i get a fine picture, more or less with some blur with the smooth motion but 3-6frames adding to the presentation glitch tally as i keep the control+J osd on for these changes.
the frame drops are quite noticeable if i leave force movie w/24refresh. which also is forced on my monitor when checked. at least as i have things configured all down the line. i use reclock sometimes and the 'treat 25 as 24fps' checked for reclock and uncheck it if i opt out of reclock to find best way to watch.
profiles are fine however way too much for all the different movies, i would have to start to make individual profiles for individual movies and then not be able to keep track of which is which since many of these old movies are from public domain and in all kinds of different formats.
I seem to have most issues with the mpeg-2 from iso
and the 'streaming video' mp4, mpeg4 , wmv are ok though the avi divx are often glitchy , too in mpc-be/mad vr.
use plenty of gpu resources , or can, if i do not alter options for each movie/video. this combo does, in most cases, show a far better final outcome than my commercial movie players from the 2 or 3major companies. i tend to use them only in last resort.
even with 6000+'presentation glitches' and not all the way through a 58 or 64min movie the picture and lip sync are excellent.
so, i am not really sure what the glitches actually are.
i read from other posts that it may be in the ffdshow and/or mpc-be components and i do check them as well along the way. only use deinterlacing sometimes, debanding rarely and disable the 'shaders' totally other than sometimes the 'greyscale' for old avi/dvix which can have a greenish tint.
just my more recent experiences since going to 87.10
Your post is written in a way that makes it hard for me to say anything. I'm not sure if you have any specific question. Or if you expect a reply, or something. If you're worried about the presentation glitches, try disabling the "present several frames in advance" options in the "windowed mode" section of the madVR settings. Does that help? Or do you then get dropped frames instead?

Quote:
Originally Posted by THX-UltraII View Post
Madshi, can you think about making a checkbox or something to disable the EXCLUSIVE notification?
This has been asked by other users, as well, but I don't accept feature requests at the moment. I'd have to add a whole new settings section just for this, and I don't consider it important at this time. There will probably a switch for this in madVR v1.0, but probably not before...
madshi is offline   Reply With Quote