Compositing Archives - Budget-stretching tips for filmmaking, animation, motion graphics & visual effects

Planning for reflections for highly efficient VFX

So this is about using additional “reflection takes” from a shoot to greatly aid in a composite.

Here’s a scenario: you shoot an actor on greenscreen that, in the final composite, will be walking past a wall made of brushed aluminum, or polished granite, or even dark plastic. You should shoot an additional take with a mirrored panel reflecting the actor mimicking the “real” take, but with blocking altered to ensure the actor never overlaps the reflection (since this will give much more leeway for manipulation in post). Then it’s relatively quick and painless to composite the reflection onto the wall or whatever, better melting the elements together visually while also giving an organic look to CGI objects without spending eleventeen thousand hours on shaders and rendering in 3D software.

Here’s some more specific guidelines…

Plan for objects that have less than perfectly crisp reflections. This is because your “reflection take” will likely not perfectly match the real take of the actor’s performance… not to mention that your reflection panel probably wasn’t a high quality mirror to begin with.
Shoot the actor’s performance on greenscreen like normal, with no reflection panel (thus allowing for a simple composite without having to roto out the panel).
Then set up a mirror or shiny board where the reflective object will be in the composite. Reblock the actor doing their performance so that you get an unobscured shot of the reflection. This will likely require some spatial cheating, but whatevs. It’ll help when compositing if the background in the reflection is either greenscreen or black (duh), but this usually isn’t super essential. I’d suggest having the actor watch the “good” take a few times so they can study & mimic their movements and timing.
In composite, finalize the actor’s position and timing into the shot. Then throw in the reflection take and tweak it to match the reflective properties of the particular surface. You’ll likely be using “Add” or “Overlay” composite modes. Btw, you should be working in linear blend mode whenever you’re doing photoreal-ish compositing or using motion blur (if using After Effects via the “Blend Colors Using 1.0 Gamma” checkbox under File > Project Settings).
Geometric mega-accuracy of the reflection’s placement and angle aren’t nearly as important in selling the shot as having the kinetic properties of the reflection match the actor… meaning that the timing of an arm swing or head turn need to be pretty close, so bust out the retiming/speed change plugin/tool if it doesn’t.

Back in ye olde 2007, this technique really saved my skin when a project I was directing/DP/compositing became logistically stressed with revisions when the signoff process grew more complicated than anyone expected. Nobody’s fault, just the nature of the beast. To put it lightly, we found ourselves in a super crunch, both in terms of time and budget. If any students are reading this, note that this kinda stuff happens all the time and probably the primary difference between a professional and amateur is that the pro is able to not bother dwelling on obstacles and just focus on figuring out a way to deal with it as best as possible within the limitations, without ripping their own hair out and going insano. Then, theoretically, you eventually matriculate to projects with big juicy budgets and long healthy schedules that are super fun with lots of creative latitude. I’ll let you know when I ever hear of one of those. The clients who give those out must hang out with Santa Claus and ride unicorns or something.

Anyways, we had to reassess a sequence that contained a CGI refrigerator. It originally was going to animate into existence in a fancy way that can only be done in 3D software, and the shot was to be 3D matchmoved. But due to aforementioned schedule/budget factors, we were hard-pressed just to get the shot done with the fridge static, un-animated. Because of the “reflection take” I was able to use just a still image render of the fridge and make the shot look “good enough”. This saved us several hours, since it nixed having to matchmove as well as eliminate the render time for the fridge. Here’s a lil’ video clip showing a breakdown where you can see how big of a difference the reflection take makes…

Is it optically accurate? Not at all. But it “feels” right to the average viewer. Is it invisible realism? Nope. But it sells the shot with a minimum amount of time and resources.

Color-saturated fill light for a better composite

Here’s a useful technique that I came up with for shooting certain kinds of greenscreen stuff…

Make the fill light a saturated color (that’s different from the screen and the subject’s clothes etc, duh).
In post, isolate that particular color and then tweak it to match the hue of whatever the natural bounce light would be in whatever environment you’re compositing them into.
Though keep in mind that it’s always better to just use a colored fill light that matches your particular background plate. This technique is more for cases where you have to use one setup for several varied backgrounds.

For example, you can use a saturated magenta for the fill, and then use the secondary color correction in Color Finesse to tweak just the saturated magenta to appear bluish if they’re supposed to be standing on the belly of a giant, naked sleeping smurf or whatever. Another great thing about this technique is that you not only can control the hue and saturation of the fill color, but also the luminance of the fill, and quite easily at that. Of course, this technique will only be useful for certain shots, depending on the lighting and blocking.

Here’s a straight forward example of this technique from a shoot I was DP on. Because of last minute schedule changes, I had to shoot numerous bits of Snoop Dogg in just one setup, even though they took place in a multitude of settings. We could’ve shot him in neutral color and depended on conventional color correction to help color him into the background, but that usually looks pretty rookie & unsavory in my opinion, since any color adjustments to affect the “bounce/ambient” light will also affect the entire subject’s color, even the areas being illuminated by the key light. So with a magenta fill, we had some tweakable “ambient” light to match the backgrounds in post.

You can see from the original footage on the right that not only was I able to easily change the hue & saturation of the magenta fill, but also the luminance. More examples from the exact same lighting setup below…

Also, note that in these examples (which I composited & color graded myself) the tweaking of the magenta fill was done with just the “Hue and Saturation” effect in After Effects due to what was available on that particular workstation. Theoretically, Color Finesse’s secondary color correction would’ve given even better results.

Animating off the (temporal) grid

Say you’re using After Effects to animate five balls flying into a wall and bouncing off. Normally, you’d have to keyframe the instance of impact on an exact frame, before changing the direction of the ball’s position path to bounce off… which wouldn’t really have any kind of qualitative disadvantage if it was just one ball. But when there’s five of them, and all of them end up having the moment of impact occur exactly on a frame, it might feel “not chaotic & organic enough”. Or just plain unrealistic if it’s a visual effects shot… if you were to shoot video of five baseballs being thrown against a brick wall, only maybe one or two would have the concise moment of impact occur exactly during a video frame. For most of them, there’d be a slur of motion blur denoting that the impact was occurring either slightly before or after an exact frame. If the balls were composited in and all had the moment of impact synch exactly to frames, then it can kinda read subconsciously as synthetic to the viewer.

This principle – that action does not naturally unfold synchronized perfectly to the occurrence of a particular framerate – can easily be applied to animation, if your project has need for such a thing. I totally just scored major snob-dawg points with that last sentence.

For sake of example, say you’re using After Effects. Put your animated layers in a precomp that has triple the framerate as your main comp… ie. if your main comp is 29.97 fps, then make your precomp 89.91 fps.
Animate the layers moving/colliding/bouncing/etc in the tripled (89.91) framerate precomp.
In your main comp, you should see that the moments of change/impact/etc are sometimes occurring at a moment in time that’s between discreet frames.
Moving any keyframes in the precomp by one or two frames allows you to put those moments of impact on or off “the temporal grid” of the main comp.
Be sure to use the phrase “temporal grid” around your coworkers to look pretentious. Err, I mean classy. Classy and sophisticated. You know, like how James Bond is.

Of course, if needed you can up the framerate of the precomp to whatever if you need more chaos/realism in the timing array of impacts.

I first came up with this when working on the project pictured below, needing smaller units of time to distribute the keyframed squares that comprise the flame movement. Since then I’ve used it on VFX-ish compositing shots for the reasons I noted above.