Correct pixel aspect ratio for a fake hd file for youtube upload

[brassplyer]

I have 4:3 DV video I'm upconverting to pillarboxed 1920 x 1080.* It's my understanding that DV video has a .912 PAR, while HD has a 1.33 PAR.

Ultimately uploading an .mp4 to YouTube.

Vegas gives the options of 1.0 or 1.33 PAR, no .912. I've tried encoding a test file with both. When done as 1.33, it looks squished played in Media Player Classic but looks correct when processed and played on YouTube.

Visually I don't see any difference between the YouTube processed videos at 1.0 or 1.33.

What's the correct way to do it? It wasn't originally shot in actual HD, how does that factor in?

Thanks for all input.

*Anticipating inquiries as to why I'm doing this, it's to retain as much quality as possible. I find that straight 480 material takes a big hit when uploaded to YouTube.

[aedipuss]

1.0 and make sure there is no check mark in the stretch video to fill output frame box. the output should have black bars on both sides.

[manono]

I find that straight 480 material takes a big hit when uploaded to YouTube.

Everything takes a big hit when uploaded to YouTube. You should devote more time to preparing your videos properly for upload. Filter them to make them more compressible so the low bitrates they use don't ruin the videos. And take care to upload a high quality 480p video to begin with.

If you're thinking your video uploaded as 1080p but played as 480p will look better than something you uploaded as 480p, that's not the case (because you'll be playing the 480p version they made from it). If you're thinking something uploaded and played as 1080p but from a 480p source will look better than something uploaded as 480p, then maybe. But then you run into streaming and buffering issues and, in my opinion, the time and effort one takes to do this isn't worth the insignificant boost in quality. I love it when people in my chosen genre trumpet something as being HD because it's been uploaded as 720p or even 1080p, when the source was obviously some crappy DVD or even a VCD or VHS tape. Like, "Just who do you're think you're kidding with that 'HD' nonsense? Oh, right, it's available in 720p so it just has to be good!"

Visually I don't see any difference between the YouTube processed videos at 1.0 or 1.33.

So you've done this already? Do you see any improvements in the 1080p versions as compared to the 480p versions from before? Be honest now. Try downloading them and figure the B/P*F to see if the 1080p versions get more bits per pixel as compared to the 480p versions. I think that's how you'd tell if the quality is better, all other factors being equal. Like you, I have read that upscaling the videos will result in better quality because the 1080p settings are better than those for 480p, even if the source is standard def. But I'm not convinced. Yes, I know the audio is slightly better on 1080p videos.

[brassplyer]

Do you see any improvements in the 1080p versions as compared to the 480p versions from before?

Absolutely.

The difference is particularly obvious when played with the intermediate and fullscreen options on YouTube. No contest between 480 video uploaded as 480 vs 480 video upscaled to 1080.

Yes, I know the audio is slightly better on 1080p videos.

That's not an insignificant consideration on video where music is a key component.

[poisondeathray]

Yes there is definitely a visible difference for youtube for upscaling HD content - if it's upscaled properly.

But I'm not convinced that 1080 is any better than 720 from a SD source . Often it looks worse because of lower relative bitrate (with respect to dimensions) and you might have those issues manono mentioned with bandwidth/ streaming . I think 720 is the sweet spot for upscaled SD content , at least for youtube at this point in time

Interestingly, I did a bunch of tests and even the 480p versions from the 720p versions get more bitrate than the 480p versions that were uploaded as 480p !? I know weird, but I tested 3 different clips and it happened all 3 times. Not enough for definitive proof, but interesting observation (it wasn't a huge difference, but still significant)

[manono]

...the 480p versions from the 720p versions get more bitrate than the 480p versions that were uploaded as 480p !?

That is interesting. And very peculiar. One would think 480p would be reencoded the same, no matter the resolution of the source.

I'm still not going to take the time using nnedi3_rpow2 to upscale and then later the extra time to upload the much larger files when my sources aren't usually all that good to begin with. But I stand corrected by both you and brassplyer and thanks.

[aedipuss]

youtube is biased to HD videos at this time. anything upped in 1080p or 720p is treated better than SD source.

[brassplyer]

I'm still not going to take the time using nnedi3_rpow2 to upscale and then later the extra time to upload the much larger files when my sources aren't usually all that good to begin with.

If it's just some generic informational video maybe it wouldn't seem that crucial but the reason I'm doing it with SD is the same as uploading real HD in an HD format - i.e. to not lose resolution any more than necessary, particularly if it's video where the visual aspect of it is important. If it's for entertainment purposes I don't see any reason not to retain as much quality as possible.

[brassplyer]

youtube is biased to HD videos at this time. anything upped in 1080p or 720p is treated better than SD source.

Ain't that always the way it is. Blatant discrimination. And this is supposed to be America!!

[_Al_]

1.DV avi from Vegas - dmfs - Avisynth (1280x720 pillarbox) - x264 encoder (CRF18) - upload to Youtube
http://youtu.be/BUT-dixMuNY

Avisource("D:\server.avi")
ConvertToYV12(interlaced=true, matrix="PC.601")
QTGMC() 
selecteven() 
nnedi3_rpow2(2, cshift="LanczosResize", fwidth=960, fheight=720)
AddBorders(160,0,160,0) #add black bars left and right to get 1280x720
ColorMatrix(mode="Rec.601->Rec.709")

2. DVavi from Vegas - denoised(neat video) - dmfs - Avisynth (1280x720 pillarbox) - x264 encoder (CRF 18) - upload to YouTube
http://youtu.be/X7jtG2JHnoU

3. Dvavi from Vegas directly out to 1280x720 - upload to Youtube (notice deinterlace (interpolate) by Vegas is not quite right)
http://youtu.be/ZqVOgteXBzw

4. DVavi from Vegas - dmfs - Avisynth (640x480) - x264 encoder (CRF 18) - upload to Youtube
http://youtu.be/j0-DFvm7pZk

Avisource("D:\server.avi")
ConvertToYV12(interlaced=true, matrix="PC.601")
QTGMC() 
selecteven() 
LanczosResize(640,480)

5. DVavi from Vegas - denoise (neat video) - dmfs - Avisynth (640x480) - x264 encoder (CRF18) - upload to YouTube
http://youtu.be/ITVjbF8eIl8

[aedipuss]

Al? What without downloading and checking all those vids is the point?

[_Al_]

the point is upload to Youtube, how it is going to end up looking

otherwise upscale makes no sense

[aedipuss]

yes and the point you were trying to make? i'm still not going to view them. what was your conclusion?

[brassplyer]

Various vids

What camera were you using? That's good SD, particularly given that we're seeing it after YouTube's downqualitying of it.

[_Al_]

yes and the point you were trying to make? i'm still not going to view them. what was your conclusion?

-that it makes sense to upscale it to get better result, if you want best result
-denoise makes not that difference at all (if footage is relatively clean), after YouTube will chew it up again, denoise can half the size of video for upload, so if this is concern why not, but encoding time will double or triple
-Avisynth upscale does better job than Vegas' upscale (interlace SD video in this case)

-and to compare those videos before uploading as I mentioned, upscale makes little sense to view it locally on PC etc. only if interlace is trouble (devices)

[_Al_]

What camera were you using? That's good SD, particularly given that we're seeing it after YouTube's downqualitying of it.

VX2000 , it was (is) good DV camcorder, I always shot manualy so there is not that much noise, but there is a lot of DV camcorders that recorded noisy picture indoor so that denoise might be worthy with those , using denoise might give better result than in this case

[brassplyer]

Here - in commemoration of the 4th of July, a video I just uploaded. It was taped with a not that great 8mm pawnshop camera a few years back. It actually didn't do too bad in daylight but video shot at night was definitely its weak point.

Epcot IllumiNations 4th of July special finale - YouTube

[brassplyer]

VX2000 , it was (is) good DV camcorder, I always shot manualy so there is not that much noise,

What do you mean when you say you shot manually?

[budwzr]

Here - in commemoration of the 4th of July, a video I just uploaded.

Thanks. You're very generous to share with us.

[_Al_]

VX2000 , it was (is) good DV camcorder, I always shot manualy so there is not that much noise,

What do you mean when you say you shot manually?

exposure is selected manually where camcorder automatically was choosing shutter speed and F-stop, or built in ND filter,
auto exposure is always too bright, sensors cannot keep up with low light and you have too much noise in low light, it is always compromise, where I was looking so picture is just right but not too dark
VX2000 had some settings where this could be adjusted though, auto-exposure, don't remember how much though,
but noise was there anyway with low light, that is just how it was, but not that bad like other consumer camcorders