1280x720 OR 1920x1080

[aruwin]

Hello!!! There are 2 videos I would like to append. But first, I have to make their resolutions the same, right?
So, which conversion is better and will produce the best quality?

Converting the 1280x720 into 1920x1080 or converting 1920x1080 into 1280x720 ??

[SameSelf]

It is doubtful your 1080 video truly has 1080 lines of resolution unless it was downscaled 2K/4K video to begin with. Therefore, you are likely better off downscaling the 1080 video to 720. This will avoid the potential of artifacts showing up in upscaled video. I will be interested in what others recommend though.

[Cornucopia]

It is a well-known phenomenon that downscaling retains apparent (though not ACTUAL) detail. Part of this is due to noise reduction due to factors similar to "supersampling".
The flaw in this assertion is this: the apparent detail only appears that detailed when viewed natively at the new rez's comparative size. So, a 1080p (with 1080p of actual detail) shown on a 1080p screen looks just as detailed, though smaller, if it is downrezzed to 720p and shown on a 720p screen that is 2.25x smaller (40" instead of 60").
If the downrezzed image is again uprezzed to display on an identical display as the original, it loses is apparent benefit and is revealed as a slightly softer version of the original. If you are doing partial frame compositions, however (such as PIP), it can still retain that apparent detail.

Uprezzing 720 to 1080 (or similar) does not have any apparent detail boost, and in fact loses a little bit of its original detail (due to interpolation inaccuracies and blending). It's just "spreading out the pixels", with a few guessed pixels added in.

So, I would say you should let the intended application/device & display (size & resolution, and DISTANCE) determine which path to take, though giving slight preferential treatment to downrezzing vs. uprezzing when all else is equal (which it usually isn't).

Scott

[ValenNC]

On a bit of a tangent, i've been debating a similar situation where I have SD footage that i'll be viewing at 1080p and my choice is, use the scalers in my video player OR take advantage of what I assume would be the superior scalers of Magic Bullet InstantHD or AE detail preserving upscaler and process the footage which I'd prefer to do anyway since they're mostly interlaced and using QTGMC would do a better job than the video player's deinterlacer.

I guess my point, and to echo Scott, either way you need to consider what resolution you'll be viewing them at and while upscaling will be just spreading pixels, wouldn't you want to use the best technology for doing that since you'll be spreading pixels either way.

[Cornucopia]

Depends on your application and the effort put into getting it there. For a compilation edit with material of multiple resolutions, it makes lots of sense to use a great (software) scaler.
For playback of consumer media files to a display, makes much less sense. Yes, you might improve the scaling (if done in software ahead of time), but don't forget you will be losing a bit of quality in the re-encode (unless you're saving to lossless). Sort of cancels out the benefit.

Scott

[_Al_]

Remember watching resampled SD on screen does not introduce one more degrading generation/ encoding.
If you resize , you must encode (with your CPU intensive method or your setup) . And it also takes more space on hardisks.

[aruwin]

Depends on your application and the effort put into getting it there. For a compilation edit with material of multiple resolutions, it makes lots of sense to use a great (software) scaler.
For playback of consumer media files to a display, makes much less sense. Yes, you might improve the scaling (if done in software ahead of time), but don't forget you will be losing a bit of quality in the re-encode (unless you're saving to lossless). Sort of cancels out the benefit.

Scott

What kind of scaler do you recommend? One that is free, of course.

[Cornucopia]

If it were me and I needed to do a high quality software scale, I'd use AVISynth's Spline36Resize, or something similar.

Scott

[aruwin]

If it were me and I needed to do a high quality software scale, I'd use AVISynth's Spline36Resize, or something similar.

Scott

But then, if I do that, I won't be able to append the 2 files to become a SINGLE video, will I?

[smrpix]

If it were me and I needed to do a high quality software scale, I'd use AVISynth's Spline36Resize, or something similar.

Scott

But then, if I do that, I won't be able to append the 2 files to become a SINGLE video, will I?

Sure you can. Uprez to match the 1080 file while re-encoding to a compatible format, and then append. Isn't that what you were asking in the first place?

[hello_hello]

On the resizing thing.....

Try creating an Avisynth script for opening the 1080p video and resize it to 720p using the script:

Spline36Resize(1280,720)

Open the script in a player such as MPC-HC and compare it to the original running in another instance of MPC-HC, both running fullscreen on your monitor (or on your TV would be better if it's connected to your PC). If you can't see a loss of detail, it's no doubt safe to downscale.
I set MPC-HC to use bicubic resizing as it's sharper than the default (the default bilinear resizing will generally make the downscaled version look softer when the player is upscaling to 1080p again on playback). Or if you don't want to rely on the player's upscaling and you're viewing the script on a 1080p monitor, you can use the script to do both the downscaling and upscaling, while comparing it to the original.

Spline36Resize(1280,720)
Spline36Resize(1920,1080)

Often, you won't be able to pick the difference when downscaling to 720p, or if you can it'll only be when comparing individual frames with lots of detail while the video is paused. Sometimes 1080p video has 1080p worth of detail and you'll loose a little by downscaling, sometimes the downscaled version can appear to have a tiny bit more detail than the original if you use a sharp resizer for downscaling and upscaling (ie spline36).

You can encode the two videos independently and append the encodes with MKVMergeGUI or MyMP4Mox GUI as long as you use the same x264 encoder settings each time (different CRF values don't matter) and you add --stitchable to the x264 command line. Of course the encodes need to be the same resolution. Possibly the same frame rate too.