How to add pillars to a 4:3 HD video stream?

[TrevorS]

I've a 1440x1080 video stream I want to display as such on a 16:9 HD TV. I've concluded the only way I can do so is to edit in graphical pillars before final authoring. Does anybody know of an HD editor that supports that?

The HD apps I'm currently working with are MultiAVCHD, AVSEditor6, and MovieMaker. Still hunting for the right combination to accomplish the job, any help appreciated :)!

================================= LATER ===============================

From searching, seems like AviSynth might be a solution, more reading to do.

[tfboy]

Hi Trevor, you can use ffmpeg to quickly re-encode the video adding pillars either side, using the pad video filter.
For example, assuming your video was an h.264:
ffmpeg -i video-in.h264 -vf pad=1920:1080:240:0 video-out.h264
where if pad=a:b:c:d, a is horizontal output res, b is vertical output res, c is horizontal offset and d is vertical offset.
As you want to center the picture, you need the horizontal offset to be half way into the added pixels, i.e. (1920-1440) / 2 = 240.

edit: if you want more control of the compression / quality of the video output, you might want to pipe the resized video into x264 or use an avisynth script. I realise what I said above might not give you the desired output quality as I haven't used ffmpeg for a video stream output, only sequential images...

[SeeMoreDigital]

I've a 1440x1080 video stream I want to display as such on a 16:9 HD TV. I've concluded the only way I can do so is to edit in graphical pillars before final authoring.Depending on the make and model of your playback device, all you should need to do is add 1:1 PAR aspect ratio signalling to your original video stream...

Provided your TV has been set-up correctly, it will add the pillars automatically.

EDIT: I forgot to ask about the container format you want to use?

[TrevorS]

Hi Trevor, you can use ffmpeg to quickly re-encode the video adding pillars either side, using the pad video filter.
For example, assuming your video was an h.264:
ffmpeg -i video-in.h264 -vf pad=1920:1080:240:0 video-out.h264
where if pad=a:b:c:d, a is horizontal output res, b is vertical output res, c is horizontal offset and d is vertical offset.
As you want to center the picture, you need the horizontal offset to be half way into the added pixels, i.e. (1920-1440) / 2 = 240.

edit: if you want more control of the compression / quality of the video output, you might want to pipe the resized video into x264 or use an avisynth script. I realise what I said above might not give you the desired output quality as I haven't used ffmpeg for a video stream output, only sequential images...
I'm actually starting with an FLV file containing 4:3 H.264. I'm trying to achieve the best quality I can, but am converting to X.264 (AVC5) to keep the peace with MultiAVCHD for final authoring (it barfs on the FLV). My thought is to adjust the 1880x1080 X264 to include the pillaring, but I'd really like to preserve the original picture quality intact.

Depending on the make and model of your playback device, all you should need to do is add 4:3 PAR aspect ratio signalling to your original video stream...

Provided your TV has been set-up correctly, it will add the pillars automatically.
Not sure what setup correctly refers to. I'm using a Panasonic NTSC 110 BD player and a current model Panasonic Viera plasma (though wouldn't want playback to be limited to them) and haven't found any menu settings that allow manually changing the HDMI HD display format (ie 16:9 Vs 4:3), no idea whether video stream commands would work.

Presuming the signalling you mention is a possibility, how does one go about accomplishing it? I'm used to basic SD-DVD authoring and have experience with PGCedit and VOBblanker, but HD and video stream mods are still new to me -- likewise the tools.

Oh, it just occurred to me you're probably talking about the settings in the player, let me check on that :)!

[setarip_old]

@TrevorS

Hi!

IIRC, multiAVCHD has a group of settings specifically for Panasonic Viera. Perhaps they meet your needs...

[TrevorS]

I've found that if I include 4:3 video in AVSeditor with 16:9 and have the project set for 16:9, then the 4:3 is preserved through adding pillars -- exactly what I want. I've also found that if a 4:3 segment is loaded into MultiAVCHD and the transcode image size exceeds 720x480 (eg 960x720), then the image can be de-cropped to 16:9 while again preserving the original AR with pillars These solutions should be independent of specific display make/model.

Only problem is AVSeditor appears to always render the input file (regardless of input/output format), and in so doing, leaves a film over the image -- yuck times two. Guess they want people to invest in their more expensive version -- my old AVC5 converter from the same company gives a nice clean render, as does MultiAVCHD. So, now I need to find a freebie HD video slicer/splicer for things I don't seem to be able to do with MultiAVCHD alone (can't see how to eliminate the chapter marker applied to the start of each segment, so need to pre-combine the segments into a single file).

EDIT: Just compared AVC5 AVI output to AVSeditor AVI output with the AVC5 file as input (same AV parameters for both apps) and found the AVC5 1929Kvbr was dropped by the AVSeditor to 1637Kvbr (both set for 3K) -- that probably explains the AVSeditor video quality drop. Am retrying the AVSeditor at 4Kvbr to see if that offsets it (don't much like it re-rendering to a lesser quality when the parameters don't require it).

EDIT2: Darn! 4Kvbr setting results in no change whatsoever in measured vbr or file size. AVSeditor just doesn't want to use the available headroom -- guess it must do an analysis and select vbr from that, additional allowance is ignored. So, it appears the only way I can match image quality between the two apps is to set AVC5 to lower than AVSeditor -- cute! FWIW -- at 3Mvbr, AVC5 is running 64.3%, and AVSeditor is running 54.6%. Seems like AVSeditor6 is not a good choice for best image quality, and their AVSconverter8 probably has the same limitation.

Final note: Tried setting 6Mvbr and again no change to the output average or file size. Just don't see any way to adjust that -- too bad :(!