[Steve Roberts]

If you choose the Displacement Map method, and the curtains alone don’t do the job, consider making a displacement map made of alternating strips, each with a gradient.

For example, the convex curtain part should be a tall strip, white fading to black on the left and right edges. The concave part would be the opposite. To make the gradient, take a strip-like solid, apply Ramp, then apply Curves and make the Curve like a parabola: an arch.

You may need to tweak the Ramps for the first two strips to get the Displacement looking right. Naturally, all these strips would be in a comp, which would then be loaded into the main comp where it becomes the Displacement Map for the text layer.

This could be a pain, squeezing each strip to match uneven folds in the curtain, but hey — you love this stuff!

[Steve Roberts]

There might be an automated way to apply a handful of pre-made (by you) Anchor Point/Scale transforms saved as presets to a lot of images … which I’m assuming is your problem. Maybe aenhancers.com has such a script.

But if you want to select a bunch of images and hit a button, that sort of thing is really the domain of home video apps such as iMovie or Windows Movie Maker. Or the slideshow feature on a DVD Authoring app.

You could also Google “Ken Burns”, “video” and “windows”. Check the Wikipedia entry there, and you may also find some downloads down the list.

[Steve Roberts]

Import the clip, and interpret the footage as lower field first. If AE doesn’t automatically do this, select file>interpret footage>main to access the dialog. When you do this, and drop the footage into a 29.97 comp, AE interpolates motion between the fields when at Best Quality, and doubles the fields when at Draft Quality.

For better quality, use something like Reelsmart FieldsKit. It interpolates the motion more intelligently. You might be also able to use Timewarp to give better interpolation by dropping the footage into a 59.94 comp and selecting 29.97, but I haven’t worked with Timewarp yet.

Really, that about it as far as I know. Each frame consists of two different half-images, neither of which exists as a complete frame. The computer has to invent the missing data somehow, usually by interpolating, hopefully in an intelligent manner.

Hope that helps …

[Steve Roberts]

Hmm … not sure I understand. If it was shot 60i, there’s no pulldown to work with, and you need to separate fields or use a plugin like fieldskit or use a method such as Andrew suggested.

If it was shot 24p and recorded with pulldown to give you 60i, then it’s a simple matter of removing pulldown on interpretation.

So. how was it shot? 60i or 24p with pulldown?

[Steve Roberts]

[Andrew Bardusk] “Then go the middle and warp that frame into place, same methodology as roto. “

Unless I’ve misunderstood, I beg to differ, old sport. Going to the middle (half, then quarter and so on) of the motion is only a useful strategy if the shape-changing motion is regular. In most cases, the motion changes at places other than half, quarter and so on, so the keyframes should be set at the frames where the motion actually changes direction or speed (points of inflection), and then AE can interpolate between them.

The trouble is, we have to watch the footage carefully to find those points. Eventually, one develops an eye for it. But it leads to fewer keyframes.

[Steve Roberts]

Yes. You access the numeric pad by pressing the Fn key and one of the tiny numbers around U,P, M, and so on.

[Steve Roberts]

… and why don’t you want to use Invigorator Pro?

[Steve Roberts]

I forgot to mention time-remapping.
This seems to work:

1. make a four-frame comp that includes the four images, one per frame
2. drop that into the main comp
3. apply time remapping to the four-frame comp, and extend it to the end of the main comp
4. make sure you have a camera (“Camera 1” in this example) in the main comp
5. apply this expression to the main comp:

thisComp.layer(“Camera 1”).transform.yRotation/2700

6. animate your camera

2700 is 90×30 to account for four images per 360, and 30 frames per second. A different number of images and different frame rate would require different numbers.
The Y rotation can go anywhere between 0 and 360. Haven’t tried anything else.

Hope that helps …

[Steve Roberts]

Oh, please do, bogiesan. It’s royalty-free. 🙂

[Steve Roberts]

Mock-wonk here … you might start with a four-frame movie containing your four views, then tie the displayed frame of that movie (a layer in the main comp) to the Y-rotation value of the camera, divided by 90 and rounded off.

Now as for specifics … maybe a true wonk will weigh in.