Rectangular Pixels in AE

[Kristan Sprague]

oh yeah, i am working on a powerbook pro (2 GHz intel core duo, 2GB 667MHz DDR2 SDRAM) in AE 6.5.1. I need to eventually render out my final comps to go back into the original project in FCP to be edited with the rest of the footage.

to give some background on the project, its a music video. the plan is the have it look like someone is looking through a box of old pictures, and the pictures are the moving images of the video. thats why i am corner pinning to index cards my talent was look through

kristan

[Steve Roberts]

Were your solids made with square or rectangular pixels?

[Kristan Sprague]

my solids are rectangular pixels. set the solids to the same specs as the footage, 720×486, D1, rectangular pixels.

kristan

[Darby Edelen]

[kristan] “do i render out at 720×486 with D1 .9 rectangular pixels? or do i render it out at 720×486 with sq pixels?”

I have to admit that for most of your post I wasn’t quite sure what your circumstances were and all of the talk of ‘pinning’ only confused me more (the corner pin effect… I think?).

But here I can be of some use. If your final output is for NTSC D1/DV then there is no difference between rendering at 0.9 PAR and rendering at Square Pixel except that other programs on your computer will be able to interpret and display the footage correctly. No matter what you set your PAR to for NTSC the output will always be 720×480 (or 720×486 for D1) square pixels. These 720×486 square pixels only appear differently with different pixel aspect ratios.

The only reason (and its a good one) that you want to set pixel aspect ratios inside of AE is so that you can preview the composition as it will appear on an NTSC monitor with rectangular pixels and the rendered footage will be interpreted properly (as 0.9 for 4×3 or 1.2 for 16×9 in NTSC) by other programs running on your computer (whose display is square pixel).

No matter how your footage is interpreted on your computer, the footage should still be 720×486 pixels. The benefit of interpreting the footage correctly and using ‘non-square’ comps with the PAR preview is that you’ll see what it will look like on a non-square display.

[Kristan Sprague]

thanks for your info, i was talking about the corner pin effect. i will try to be more clear from here on in my problems.

here is my workflow so far.

I place a video clip (NTSC, D1 720×486, .9 PAR) in a comp. It is a point of view shot, all you see is someone’s hands looking through a stack of colored, blank index cards. I then place a solid layer above that video clip, (same setting). Using the corner pin effect, I go through the comp frame by frame, “pinning” each corner of the solid to the corners of a moving index card. when i am done, i watch through it to ensure that the second layer, with the pin effect, follows the card it is pinned to as the talent moves it around.

I then remove the video layer, the guide layer if you will, so that all i have is a solid that moves around in the exact same path as the card it was covering. I then add a different video layer underneath that layer. again, this video layer is NTSC D1 720×486 with rectangular pixels. I set the solid layer to an adjustment layer so that the pin effect i rotoscoped now has the exact same pin effect applied to it. Up until here, i have been doing fine.

it’s when i then try and nest that composition into another comp that contains the background i want to composite the corner pinned image to. I reuse the same video layer that i used to guide my corner pin in a new comp with the exact same setting. so theoretically i should see my corner pinned footage over the background footage, both moving in lock step with each other. i do not see this. the nest comp with the corner pinned footage seems to move around in the same path, but it’s size is always off. they never match up exactly. if i copy and past the solid layer into another comp, and view it as a regular layer not an adjustment layer, it works fin. it is only when i precompose a comp that uses the solid as an adjustment layer that everything goes out of whack. the ‘adjusted” footage, not the solid, is always slightly smaller than the original solid it uses as a guide layer, even if i increase the scale on the it while leaving the solid the same size.

could this be because i am viewing rectangular pixels as square pixels? do the corners that i am pinning to the card only appear to be pinned exactly at the corners, but in reality are slightly off because of the different PARS and views?

“The benefit of interpreting the footage correctly and using ‘non-square’ comps with the PAR preview is that you’ll see what it will look like on a non-square display.”
I understand that rectangular pixels in a square pixel display will “appear” squeezed, but in reality they are not. so if i track a corner of a index card manually, frame by frame, placing a solid object at that exact point in each frame, am i really exactly at that point, or does it only “appear” that i am exactly at that point because the image is squeezed?

or more simply, to do what i want to do, should i be using 720×540 comps with sq pixels, 720×540 comps with D1 rectangular pixels, 720×486 comps with D1 pixels, or the NTSC D1 preset AE has where my frame size is 720×540, my pixels are square, but their PAR is .9 for D1?
do i need to set my PAR preview to rectangular pics when i corner pin a layer and as i edit my comps?
when i am ready to render out for FCP, do i simply render the comps with the settings i edited in, or do i need to move the contents of those comps to new comps set to 720×486? and if so, do i set the PAR in these comps to square or to rectangular?

kristan

[Darby Edelen]

[kristan] “I understand that rectangular pixels in a square pixel display will “appear” squeezed, but in reality they are not. so if i track a corner of a index card manually, frame by frame, placing a solid object at that exact point in each frame, am i really exactly at that point, or does it only “appear” that i am exactly at that point because the image is squeezed?”

Just to clarify further, the ‘rectangular pixels’ on a square pixel display will not appear squeezed, they will appear stretched/scaled horizontally by 111%. Square pixels will appear squeezed on a 0.9 PAR display down to 90% horizontally.

In order to display the rectangular pixel video as it will appear on a 0.9 PAR device AE scales the 0.9 PAR footage (appears to be stretched 111.111…%) down by 90% (111.111…% * 90% = 99.999…% or effectively 100%).

Now, on to your specific situation =) If you want the best accuracy with placing the corners you should turn off the PAR preview for the comp when you’re placing the corners, this won’t have a huge effect (and is probably not the source of your trouble) but does give you the best view of the actual pixels on a square pixel display.

As far as the mismatch you’re seeing in size, it sounds like one of your compositions might be using square pixels. I’d recommend going back in and making sure that all of your footage and all of your compositions are set to the NTSC PAR of 0.9.

I don’t work in NTSC D1 (720×486), I work in NTSC DV (720×480). I always use the NTSC DV (or DV Widescreen) preset for my compositions to ensure that I have a uniform 0.9 (or 1.2) PAR for every comp.

I never work in ‘square pixel’ comps unless my render is going to be viewed only on a computer’s square pixel display.

[Kristan Sprague]

thank you again for the info.

just want to ensure i’m taking your advice correctly though:

there are two presets for D1 comps. one with square pixels that stretches the frame to 720×540. the other uses D1 rectangular pixels and the regular 720×486 size.

you’re saying i should use the regular D1 preset, with a 720×486 frame and D1 .9 Par, correct?

what if i use square pixel assets in the comp? say a scanned still picture? do i keep the comp at .9 PAR and 720×486? or if i;m going to mix square and rectangular pixels, what setting do i use for my comps? the .9 D1 PAR at 720×540? and if so, do i then need to render that content to a 720×486 comp with rectangular pixels?

it seems like the deeper i get into this, the more confusing it gets.

kristan

[Darby Edelen]

[kristan] “what if i use square pixel assets in the comp? say a scanned still picture? do i keep the comp at .9 PAR and 720×486? or if i;m going to mix square and rectangular pixels, what setting do i use for my comps? the .9 D1 PAR at 720×540? and if so, do i then need to render that content to a 720×486 comp with rectangular pixels?”

If you’re creating assets yourself you should create them at 0.9 PAR just to be safe. If you’re scanning an image in then the asset should be interpreted as Square Pixel (or its native PAR, whatever PAR it looks correct at) in AE but your comp should still be 0.9 PAR.

As long as your assets are interpreted at their correct PAR and your comp is set to the right PAR for your footage’s final destination, AE should handle most of the dirty work.

For example, you could create a square pixel Photoshop document, draw a circle in it and import it into a 0.9 AE Comp and it will still appear as a circle as long as the imported PS doc is interpreted as square pixel footage (the PAR it was created in).

All of this assumes that you have PAR preview enabled for your comp (I always do).

[Accountneedsrealnameupdate]

Corner pinning does have some gotchas that drive me nuts sometimes (Layer co-ords vs Comp co-ords is the big one). Your description is pretty good but even still I’m not 100% certain of what exactly the problem is do I’m going to suggest my own fix when I get this stuff wrong.

I would recommend leaving the corner-pinned elements in the comp you built them, turn off the background layer you used to track the corners into place and then nest that whole comp into the 486 comp with the original footage, if it’s not already filling the whole comp without spilling over the edges (if all your PAR’s are right it should fit perfectly) use Cmd+Opt+F to “fit to fill”, usually that just locks everything back in the right place and corrects mistakes with PAR, almost like corner pinning it back in without the hassle of motion. (Don’t collapse transformations though as it will mess it up again
)
Good luck, I’ve had some hellish weeks getting corner pinning to look nice, if it’s off at all you get a horrible wiggle effect.
Glennser

PS the advice so far was spot on, I always find with PAR stuff once you get the interpretation right After Effects will look after the rest, when in doubt fit to fill usually fixes it for me though.