MKV: Frame Sequential using two video tracks

[an3k]

Hi,

i want to mux a 3D H.264 1080p video using matroska. I hate Side-By-Side because you lose horizontal resolution and Interlaced because ... well .. interlaced so i want to use the Frame Sequential 3D Format (1st Frame for the left eye is taken from track1frame1, 1st frame for the right eye is taken from track2frame1, and so on).

My question now is if i have to manually re-encode both video tracks into one or if i can tell matroska to simply use 1st frame from video track 1, 2nd frame from video track 2, third frame from video track 1, etc.

Thanks

[sneaker]

While Matroska should support that, I fear there are currently neither muxers nor demuxers capable of using it.
http://matroska.org/technical/specs/notes.html#3D

So at the moment you're stuck to using a single track.
But:
1.) You don't have to stay in 1920x1080, so there's no rule saying you have to lose horizontal resolution
2.) If you absolutely want to stay in 1920x1080, you may want to use top/bottom. Since many movies use something like cinemascope, you may be able to keep more resolution.

[an3k]

Thanks for the information! I thought this is already possible since it's like a zipper = very easy

I found that blog entry (http://mod16.org/hurfdurf/?p=8) some time ago and thought about testing Frame Sequential using a virtual timeline build with chapters. I don't know how much more work the demuxer has if it has to switch the video track after every frame but haali could handle that. The only problem will be that you always have to tell your 3D display if the current frame is for the left or the right eye, especially every time after seeking.

I also could recode both video streams into one and would have the same advantages/disadvantages but then the "3D" video is not backwards compatible (same with SBS and TBB) - you always have the useless right eye part on your 2D display.

[sneaker]

Thanks for the information! I thought this is already possible since it's like a zipper = very easy

It probably really is relatively easy to implement, it's just that no one has bothered to do it yet. (AFAIK)

I found that blog entry (http://mod16.org/hurfdurf/?p=8) some time ago and thought about testing Frame Sequential using a virtual timeline build with chapters. I don't know how much more work the demuxer has if it has to switch the video track after every frame but haali could handle that. The only problem will be that you always have to tell your 3D display if the current frame is for the left or the right eye, especially every time after seeking.

That's just insane. I'd drop that idea ASAP.