Non-square pixels, keeping 4:3? (Sony Hi8/DV)
 

Hello!

I have a question regarding the workflow for footage from Sony camcorders (PAL Video8/Hi8/DV).
So I bought an old, but quite high spec Windows XP machine just to transfer camcorder footage via firewire, and I copy the tapes with WinDV and it's working like a charm with my DV cameras.

I would like to know what's the best approach to handling this footage (720x576). I mostly work in Premiere Pro and After Effects for video work (I also do Davinici Resolve as I'm also shooting on BMD cameras, but I'm so used to Adobes suite).

Now, inside of Premiere, I can just select the pixel aspect ratio to D1/DV PAL (1.0940), and it translates to a 4:3 sequence, perfect. My issue is when I want to render this out, some platforms does not know what to do when playing the video. Let's take Instagram as an example, since they squeeze the video into 5:4 (it looks like that, squashed from the sides).

Would you rather have the timeline as 720x540 px, square pixel ratio and de-squeeze the footage with this factor? Or keep it as a non square pixel ratio timeline and then sort it out in the rendering/encoding stage?

What is the professional approach? I'm here to learn! :))

Best,
psykel

Sort it out in post. If you want 4:3, you sometimes have to force it. De-interlace and rescale to your output's target resolution. Add pillar boxes to pad out the sides. A common workflow dealing with YouTube (because 480p looks terrible and only 30fps) is to de-interlace to 50 or 60fps, resize to 720p and add pillar boxes.

Square pixels all the way for me! The only other main-stream non-square video format is HDV (1440x1080i 16:9- niche from HDV camcorders). Everything else is square (DVDs will probably be non-square but that is a special-use case). Why not make your SD videos square pixel as well. No issues then with sizing/playing on any device/player anywhere.

I have never added black bars for YT. If you do, YT will display them. If you want to resize 4:3 SD for YT, just do it: 960x720. If you "add pillar boxes to pad out the sides" you're retaining the squashed 5:4 shape of the actual video.

Personally, I like the look of 480p at 25 fps even if I get that people enjoy higher frame rates for sports (I play Counterstrike with 400 fps, so I know). But, I do think SD at 25 frames is a nice look when it's chosen deliberately. Or do you mean that in this specific case, Youtube encodes 480p videos a lot worse than 720p? I'd rather stick with the original resolution to be honest! But I'm looking into proper ways to deinterlace the footage for web.

Almost everywhere I look, people seem to mention h264 as their go-to encoding--is this still the case. When I render things and need to keep a "quality" render I usually do a ProRes 422, and h265 for web (used to do h264 up until this summer or so). Is it h264 lossless that's the thing or what is the appeal over other encoding options? I'm have 500 tabs open in my browser, trying to install VapourSynth but an old, incorrectly wiped anaconda3 install doesn't allow me to proceed. -_-

Thanks for the reply!

So would the best approach for dealing with my 4:3 VX/PD DV PAL footage, to (in Premiere) have the sequence settings set to 720x540px, with a square pixel ratio, then de-squeeze the footage with the proper factor of 1.0940 (found a video: https://youtu.be/IHFN3yyrf98)?

And since everything is online, it would make sense to make deinterlaced versions of the copied tapes for online use I take it? I have my toes in most things media, art and music so I'm super spread out and it's moving slowly on all fronts, but I'm eager to learn, always! :)

I don't use Premiere but the 720x540 will give you 4:3. I'm sure that YT video is relevant here: he's talking about stretching anamorphic video from a drone or action cam. I don't think that applies to your 4:3 DV. In any case, I think you're doing a similar thing by putting your video into the 720x540 project and filling the screen. I don't understand the 1.09 bit.

Any Premiere Pro users help here?

If social media is a potential target de-interlace and resize to 1440x1080 1:1 SAR for 4:3. Don't bother with any intermediate resolution, will be waste of time and quality.

I'm assuming final footage is not intended for SD.

I would QTGMC deinterlace 720x576 in Avisynth first, to 50fps (p50, aka 720x576p50), and to a lossless codec that Premiere can see. I'd probably opt for Lagarith for this non-capture intermediary.

In that Avisynth script, after deinterlace, I'd crop to 704x576, top remove pillar bars.
Then resize to 720x540.

If needed, then mask any overscan, or crop by 4:3 (for every 3 from top/bottom remove, 4 from left/right must be removed). The resize again to 720x540 if cropped, as cropping reduced size some. Use a good resizer, and your footage will determine which is needed. Bicubic safest. Yes, resize + resize again will cause a % of loss, but it's easier to screw up the aspect-accurate cropping by trying to crop in 5:4 720x576 domain.

Then create a Premiere project that is 720x540p50, and enjoy.
After Effects, enjoy, fun.

For streaming/Vimeo/Youtube, I'd resize to 1440x1080p50.

If required, only then add pillarbars from 1440>1920. I generally do anyway, preference to adhere to 1920x1080p50 for others uses.

"More than one way to skin a cat." (Poor kitty. :( )

Yeah that's often the case these days, so that makes sense, thank you!

Thank you so much for this huge chunk of information! I will do my best to look at documentation and guides for getting used to the softwares, and with this step-by-step guide, I'm sure I'll figure it out! I have filmed stuff (for fun) for 20 years, and I work with making video, but in this SD realm, I have so much to learn, and I'm making progress thanks to communities like this one. When SD footage was the thing I just played around with it like kids and teens do, and if it worked, it kind of worked. But now I really want to get the most out of these formats. :)

Thanks again everyone, I'll dive back into these SD waters with all the new information I've been given! :salute:

To add, to avoid confusion, to reiterate:

- Cropping 720 > 704 is important to main the 4:3 AR, prior to resize.
- Then you can resize.
- You can also mask/crop, but must do so with a 4:3 to again not mess up AR.

Most of the crop/mask is about AR integrity, essentially to not make people skinny or fat, to not make circles into ovals.

The pixel AR is really secondary to it, but I hate doubling up my AR considerations. Yes, it may reduce quality by a slim %, but I don't need a headache from having to triple check math formulas.