MPC: Override aspect ratio

[mikeggsm]

Hi,

I have a video with the wrong aspect ratio, it should be displayed in 4:3 but it's encoded stretched/distorted to 16:9 (people look short and fat).

In MPC-HC 1.5.1.2903, right Click -> Video Frame -> Override Aspect Ratio -> 4:3 doesn't change anything... why?

Thanks.

[mikeggsm]

It works perfectly in VLC... don't force me to use it please

Nobody knows why this seemingly basic function doesn't work, am I doing something wrong?

[JanWillem32]

For writing an aspect ratio on a video stream header that has the wrong properties, MKVToolnix' MKVmergeGUI is convenient to use. http://www.bunkus.org/videotools/mkvtoolnix/
(If the file is already MKV, you can also have a go with MKVpropedit in the command line interface to edit a file directly.)

For using the player's own aspect ratio controls, it's often enough to disable/enable "Keep Aspect Ratio" under the "View", "Video Frame" menu.
You can also make a preset Aspect ratio correction.
These can be made by choosing: "View", "Pan&Scan", "Edit...", "New".
In your case, vertical stretching needs a zoom factor of 1 to 1.33333·, horizontal shrinking needs a zoom factor of .75 to 1.

[mikeggsm]

JanWillem32: Thanks but nothing was working, Keep Aspect Ratio checked or not, and Pan&Scan wasn't doing anything.

Actually I just found out why: it's because I use the renderer EVR, not EVR CP. With EVR CP, Override Aspect Ratio to 4:3 and Keep Aspect Ratio to "checked" works great.

Except that, as usual with EVR CP, the screenshots don't take into account any viewing aspect ratio (neither AR encoded in the file nor the AR under Video Frame -> Override Aspect Ratio), right now they look distorted as in the original video, not 4:3 as displayed on the monitor.

I can always hit Printscreen though.

[JanWillem32]

The screenshot function needs updating. It's slow, can't fully save an image from a renderer stage and has no support for the internal resizers. I've already proposed a new function, but right now, the print screen method does indeed produce the best quality pictures. Standard EVR and VMR-9 windowed both use internal methods to render a picture, there's only very little control on those.

[mikeggsm]

I agree, I posted about it last month http://forum.doom9.org/showthread.php?t=161632

The problem with Printscreen, in addition to being inconvenient (need to paste it in a different program and save it by hand), is that it saves the black bars in fullscreen, and they can cause problems in many situations.

One strange thing I have noticed many times in EVR is this. Make the video full screen, then hit the Save Image hotkey. The image's quality is OK, but it's not exactly what's on the screen, it has artifacts like horizontal bars (printscreen gives better quality, exactly what's on the screen). However the black bars are saved too.

To avoid saving the black bars, go back to windowed mode (the window fits the video's original definition), Save Image, and then view that image in fullscreen with any viewing program (eg FastPictureViewer). That image looks really blurry compared to the fullscreen shot.

I avoid that problem by still saving in windowed mode but making the MPC window very big (with ctrl + resize window, to keep the AR of the video).

It looks like, when the video is decompressed into a smaller resolution like the original video's resolution, and then the resulting flat image is later resized into a larger resolution, compressed information stored in the file (intraframe DCT tables or interframe motion information I suppose) is wasted, compared to directly decompressing that information into the larger resolution.

It is not surprising since the limit case where you decompress the video into a 10x10 pixels image and then resize that flat image to fullscreen will obviously lead to a terribly blurry image; but what's a bit surprising at first is that the argument remains true when decompressing into the video's original definition: decompressing to fullscreen still leads to significantly better quality compared to resizing the original definition image to fullscreen, so decompressing to the file's original definition does not use the full compressed information encoded in the file.

EDIT: if your new function works from EVR-CP, gives Printscreen quality without the horizontal artifacts and works with the file's display aspect ratio, then it's perfect and it would be great if it could be incorporated quickly...