My first question is why are you creating an anamorphic composition. The standard is Square Pixels and standard frame sizes. If you are using something like DVCPro HD as a preset for the comp then the only possible use for your footage would be to save it back to a DVCPro HD player through a capture card. Nonsquare rendered footage are not a good idea for web, mobile, or computer playback and generally won't work. They may be kind of usefull for creating a DVD, but not for a BluRay disk. Even with those delivery options square pixels are better except for the final compression to build a DVD.
Second question, or rather second suggestion, why don't you send us a screenshot of your project with the properties of all layers revealed. This may provide a clue. Scaling down will soften an image but it should not make it look pixelated.
Thanks for replying Rick. The base footage I'm working with was shot anamorphically on an Arri Alexa. So in essence, I have an anamorphic "bg plate" and want to comp in a non-anamorphic "fg plate."
I've read that when you toggle the pixel aspect ratio correction in the comp view it can make elements appear a little pixelated, but it shouldn't be that way once you export. But that's not what's happening.
I may have found a workaround by treating the anamorphic footage as square pixel footage--meaning creating a comp at the 2:1 width and then stretching the footage out. That seems to alleviate the problem, but I'm not sure if I'll have any quality loss doing it this way.
Here's a little screen grab of the anamorphic comp at 100%:
Here's the same piece in a square comp:
Here's what is happening. The square pixel exr is being stretched and re-sampled in your non-square comp. Changing the comp to square pixels removes that problem but it introduces distortion to your image that will show up when the shot is projected. You would get the same result if you kept the comp rectangular pixels and changed the interpretation of your exr images to non-square the same as the comp. The problem with your workflow is that circles will become ovals, lines at 5degree angles will be at 7, thin lines will become aliased.
IOW, in the long run you did nothing to solve your problem because this very thin white line is going to be stretched and re-interpreted when your project is completed as a square pixel deliverable for projection or distribution in film.
I've been doing this for years and the best result most of the time is to work all in square pixels and deliver in square pixels. Camera footage will almost always look better if it is placed in a square pixel comp than graphics or still images placed in a non square pixel comp. Run some tests and check the final output of both workflows. You only need to render a single frame to see the difference.
OK, so if I understand you correctly, when working with footage that was shot anamorphically and is squeezed at 2:1, you don't use an anamorphic comp setting or interpret the footage as anamorphic? AE automatically interprets the footage as 2:1 and creates an anamorphic comp when I create a new one with the footage.
If I change the comp and footage settings to square pixels I have to stretch the footage's width by 200% to make it fit, so isn't that the same as working in an anamorphic comp? I would then have to re-squeeze the comp for delivery, which AE automatically does if i'm working in an anamorphic comp.
Also, I put a circular shape layer in my anamorphic comp and it looked just fine. It becomes an oval when re-squeezed for delivery, but when unsqueezed for projection it would look fine.
I think what's going on here is that you have Pixel Aspect Ratio Correction turned on and that's why you're seeing the jaggies in your anamorphic comp. If you turn it off, you will see that the pixels being rendered by AE are nice a smooth as you'd expect, although the image will be squished horizonatally. You should work with Pixel Aspect Ratio Correction turned on, but know that the jaggies are an artifact of the drawing.
When you put the footage in a square comp, AE resamples the image using bilinear filtering, so the image looks better. Pixel Aspect Ratio Correction is done on the fly so it uses the coarser nearest neighbor filtering. Maybe a good feature request would be to have it use bilinear filtering instead at the expense of a little delay in drawing. If done on the graphics card, there might actually not be any noticeable delay.
How are you going to deliver your final result? If it's going to be anamorphic like the plate is, then you should set all your comps to anamorphic and let AE do the math for you.