Fixing (not upscaling) old VHS-Camera shot

[Sharc]

For testing i took 1150x1053 (as described in Selurs link for analog sources), but it is not the same as in source from post #1. But actually im not sure which one of them is the right one. Actually quite difficult to tell.

Is 1150x1053 common for the VHS-8 PAL tapes ?

PAL VHS capture is 720x576 with small black borders left and right. Anything else is the result of some subsequent rescaling by your upscaling settings in Hybrid. I wouldn't upscale at all.

If you want to convert to square pixels set the input PAR 12/11 and the output PAR 1/1, or set the input PAR 12/11 and set the output frame size 786x576.

If you crop the source to 704x576 (removing the black side bars) set the input PAR 12/11 and the output PAR 1/1, or set the input PAR 12/11 and the output frame size 768x576 (or any other 4:3 ratio like 1440x1080 if you want to upscale).

[Platos]

For testing i took 1150x1053 (as described in Selurs link for analog sources), but it is not the same as in source from post #1. But actually im not sure which one of them is the right one. Actually quite difficult to tell.

Is 1150x1053 common for the VHS-8 PAL tapes ?

PAL VHS capture is 720x576 with small black borders left and right. Anything else is the result of some subsequent rescaling by your upscaling settings in Hybrid. I wouldn't upscale at all.

If you want to convert to square pixels set the input PAR 12/11 and the output PAR 1/1, or set the input PAR 12/11 and set the output frame size 786x576.

If you crop the source to 704x576 (removing the black side bars) set the input PAR 12/11 and the output PAR 1/1, or set the input PAR 12/11 and the output frame size 768x576.

Ok, thanks. But how do you know which PAR ratio is the right one? Because the one source from #1 seems to have 16/15 (hybrid display that as input). And with 1150x1053 i meant PAR from this side: https://forum.selur.net/thread-597.html
I do not upscale this source.

Because when i use input 12/11 and output 1/1 i get sth little bit less wide than the source from #1.
But as i said im not sure anymore what actually is the "right" stretch here.

Sorry when i ask so much, but im not sure anymore what stretch is the real one. Dont want to produce videos which are less stretched than "original". But the source from #1 is not that good so i can also imagine the one who did this source did stretch it to wide.

[Sharc]

Ok, thanks. But how do you know which PAR ratio is the right one? Because the one source from #1 seems to have 16/15 (hybrid display that as input). And with 1150x1053 i meant PAR from this side: https://forum.selur.net/thread-597.html
I do not upscale this source.

Oh I see what you mean. Your notation 1150x1053 confused me. I misunderstood it as your upsized frame resolution.
The PAR notation is 1150/1053. Now there are slightly different approximations for the PAR, the reason behind this is a bit involved, so here just in short what you might occasionally find:
a) 12/11 = 1.090909.... Origin: mpeg4 standard, digital realm
b) 1150/1053 = 1.09212.. Origin: ITU/Rec601 standard, for analog sampling
c) 59/54 = 1.09259 derived from the sampling ratio 14.75MHz/13.5MHz=59/54, used by Vdub for example.
These ratios are all very very close and within pixel accuracy. The slight differences are of little practical relevance. Analog video has no pixels, and the x/y notations are approximations. b) and c) are most accurate for analog video captures, a) is easier to keep in mind and is perhaps a bit more "practical", to say it a bit sloppy.
So nothing to really worry about. Use any of these.

Note that above applies for the capturing (sampling) of analog footage (like your VHS tapes) acc. PAL.

Edit: The 16/15=1.066666.... of your post#1 would apply for mpeg4 (digital realm, DVD) with a picture filling the full 720x576 area, i.e. without any black side borders. It is actually incorrect for analog captures, resulting in a slightly distorted picture (2.4% error, squashed horizontally). Casual viewers would usually not notice this small error though.
In your case the active picture within the 720x576 frame is ~702x576 and needs to be stretched according to the PAR listed above which will give a correct 4:3 frame for the active (without the sidebars) picture.

All this in a summary / rule of thumb:
Crop your capture to 704x576 and encode with a PAR of 12/11, or crop to 704x576 and resize to 768x576 (or any other 4:3 ratio like 1440x1080, after deinterlacing) for square pixels.

[Platos]

Thanks for your further help.

So then i believe i am doing sth wrong, because when i use 1150/1053 in hybrid, i get a big difference compared to the one source from #1. I attached 2 pictures. The wider one is from #1, the other is from 1150/1053.

So can you say me, if that is the "slight difference" or am i doing sth wrong here? Its a screenshot which i made smaller because of filesize limitation.

[Gelinox]

And without Avisynth ?

[Selur]

if your source is wrongly flagged, you should overwrite the input par

[Platos]

And without Avisynth ?

Sorry, i dont understand what you mean. I use vapoursynth in hybrid for deinterlacing. So it should be no avisynth i think.

[Platos]

if your source is wrongly flagged, you should overwrite the input par

Ahh, but how do i know if it is wrongly flagged and even more important: how do i now which is the correct one?

Should i better use fix Resolution ? Problem is i would have to calculate because of the cropping i guess.
Edit: Testes fix resolution without cropping and it does not work either.

[Sharc]

Thanks for your further help.

So then i believe i am doing sth wrong, because when i use 1150/1053 in hybrid, i get a big difference compared to the one source from #1. I attached 2 pictures. The wider one is from #1, the other is from 1150/1053.

So can you say me, if that is the "slight difference" or am i doing sth wrong here? Its a screenshot which i made smaller because of filesize limitation.

As explained before:
The input PAR for #1 (the .mkv file) is 12/11 (or 1150/1053 if you prefer), not 1:1 or 16/15
The input PAR for the elagato capture is 12/11 (or 1150/1053 if you prefer)

And you have to decide whether you want to keep it anamorphic (non-square pixels) or convert it to square pixels.

[Platos]

As explained before:
The input PAR for #1 (the .mkv file) is 16/15, not 1:1
The input PAR for the elagato capture is 12/11.

Im super confused now

So i should use 12/11 as input and oputput ? But why? The elgato-source is 720x576 and the aspect ratio is 5:4, so the PAR should be 1/1 and not 12/11 ?

And why makes cropping the immage a difference ? When i use 12/11 as input and output with no cropping it is not the same strecht like when i use a 700x568 crop. I mean shouldn't the PAR be the same after a crop?

Because of the ugly line on the bottom of the video i have to cut off 8 pixels.

This is so confusing

[Sharc]

As explained before:
The input PAR for #1 (the .mkv file) is 16/15, not 1:1
The input PAR for the elagato capture is 12/11.

Im super confused now

So i should use 12/11 as input and oputput ? But why? The elgato-source is 720x576 and the aspect ratio is 5:4, so the PAR should be 1/1 and not 12/11 ?

And why makes cropping the immage a difference ? When i use 12/11 as input and output with no cropping it is not the same strecht like when i use a 700x568 crop. I mean shouldn't the PAR be the same after a crop?

Because of the ugly line on the bottom of the video i have to cut off 8 pixels.

This is so confusing

Correct, cropping does not change the PAR.
Cropping or padding chages the DAR but keeps the PAR intact, resizing (changing the width or hight) changes the PAR.
Sit back and relax, and forget the comparison with the crappy #1 .mkv

If it helps:
Take your elgato_cut_VideoOnly.avi capture which has a PAR of 1150/1053 (~12/11)
-If you crop it to 700x568 and want square pixels the output frame will be 764x568
-If you crop it to 700x568 and want to keep it anamorphic like the source leave it at 700x568
You'll find your way how to do this in Hybrid.

[Platos]

Ok, i actually do not care about the source from #1. I only want, that the video is stretched as correctly as possible without the need of a player who fix it.

So actually i would be glad if someone could tell me, if these hybrid-settings are the right for it:

Especially the "auto adjust" on crop/resize->resize ist sth, which drives me crazy. Do i have to uncheck that while changing the PAR or do i have to set it on auto ?

and if this setting is the right one: Can i now crop the picture without changing sth else without changing the right "stretch" or do i have to change sth then?

Sorry to all, but i would be very glad, if someone could tell me, what is correct.

[Sharc]

Ok, i actually do not care about the source from #1. I only want, that the video is stretched as correctly as possible without the need of a player who fix it.

OK, so you want to convert it to square pixels.
For example like this with your cropping and keeping the vertical resolution intact. You see that the output resolution will become 764x568 as I wrote before.

So again: The PAR of your capture is the Input PAR which is 1150/1053. As you want to convert to square pixels you select the Output PAR as 1/1. Now you can play with with the Auto adjust target resolution. You can now set the width OR the height but not both, because one depends on the other to fullfill the square pixel condition. Hybrid is very smart.

[Platos]

ok thanks. I do it now this way, thanks alot.

[Sharc]

As you may readily conclude the 764x568 output is no longer exactly 4:3, but 1.34507. The reason is of course your cropping of the source. But the objects are displayed correctly (=undistorted), means a perfect circle will still be a perfect circle when played back in square pixel view.

If you wanted an exact 4:3 overall frame you would have to pad the picture (=adding black borders) accordingly, so for your cropped example the overall frame size becomes including the borders 768x576 (=4:3), square pixels.

[Platos]

Yeah, that's what i wanted. Displayed correctly everywhere, also on devices which does not correct the wrong stretch. Therefore i also save it on h.264 for maximum compatibility and .mp4 (not .avi or .mkv).

I have to remember that with the cyrcle. That's the easiest way to describe what i want with all these different Stuff (PAR, DAR and so on). Maybe also a good "test". Search sth you know it should be a circle. If it is not one anymore it is wrong stretch. But for sure often there are no circle-like things.

[rgr]

Edit: The 16/15=1.066666.... of your post#1 would apply for mpeg4 (digital realm, DVD) with a picture filling the full 720x576 area, i.e. without any black side borders. It is actually incorrect for analog captures, resulting in a slightly distorted picture (2.4% error, squashed horizontally). Casual viewers would usually not notice this small error though.

I checked what PAR was generated by DVD recorders and HDMI grabbers and it was always 16:15 (either 704x576 (DMR ES10) or 720x576).
Well, maybe one day I'll find a cassette with something perfectly round, I'll measure it

[Sharc]

Rec601 specifies a luma sampling rate of 13.5 MHz which for PAL gives a Pixel Aspect Ratio (PAR) of 59/54 aka 1150/1053, approximated by 12/11. Therefore the active picture width of a capture within a 720 wide frame is ~702 pixels, and the 4:3 picture is represented by the inner ~702x576 pixels.

DVD is a digital standard which does not even specify the PAR but only the DAR (like 4:3, 16:9). Being a digital medium it follows the mpeg2 standard which violates Rec601. The implicit PAR of a 4:3 DVD (including the sidebars of captured analog footage) then becomes 4/3*576/720=1.0666... = 16/15, and this is what tools usually report, although the PAR of the captured inner ~702x576 picture is actually acc. Rec601 (approximated by 12/11 for 4:3 PAL).
Commercial DVD content may however be PAR 16/15 or 12/11, it's not specified, and one may only find out with the circle test.
A DVD will always be reported as 16/15 and played back as 4:3, irrespective whether its actual video has a PAR of 12/11 or 16/15. DVD is just sloppy.

Commercial VHS tapes may sometimes follow their own rules I am afraid, especially as we don't know on what master footage these are based on.

[The_Doman]

Yeah, that's what i wanted. Displayed correctly everywhere, also on devices which does not correct the wrong stretch. Therefore i also save it on h.264 for maximum compatibility and .mp4 (not .avi or .mkv).
I have to remember that with the cyrcle. That's the easiest way to describe what i want with all these different Stuff (PAR, DAR and so on). Maybe also a good "test". Search sth you know it should be a circle. If it is not one anymore it is wrong stretch. But for sure often there are no circle-like things.

I checked what PAR was generated by DVD recorders and HDMI grabbers and it was always 16:15 (either 704x576 (DMR ES10) or 720x576).
Well, maybe one day I'll find a cassette with something perfectly round, I'll measure it

Yeah, i was at one moment struggling too with this issue when playing the video's on PC/TV.
You need a good visual reference to see what the player/TV/display is doing exactly during the playback.
I found some old VHS recordings (not discussing quality now) of testcards which helps configuring/understanding it correctly.

This well known (PAL) type pattern works best i think.

[Attachment 78850 - Click to enlarge]
Captured DV 720x576 format, should play correctly in 4:3
Converted MP4 720x576 format with set metadata for 4:3, which should then play correctly on media-players and TV's without the need for the extra horizontal resizing step.

I used AVIDemux here:

[Attachment 78854 - Click to enlarge]

[Sharc]

The "circle" in your screenshot is about 504 pixels high and 462 pixels wide. Hence its estimated PAR is 504/462 which is 12/11=1.0909....
You have to stretch the picture's width by 12/11 to get the circle correctly back as a circle.
Including the black sidebars the full image is then 12/11x720/576=785/576=1.36:1.

[rgr]

If VHSDV_BRTN_Testcard.avi were written with PAR 12:11, it would be perfectly round. Recorded with PAR 16:15 it is 2% too narrow (497 wide, 505 high).
The second one, of course, completely bad.