[Update] Perfect Parking Spot – Shot 1-3: Rendering and Compositing (Patreon)

Finished up work on rendering and compositing the pre-stomp build up shots. Diving into some more compositing stuff this update, ahead of getting into the complex scene-building and simulations stuff the remaining post-stomp shots require.

Breaking up shots into layers for compositing

I haven't been diving into too much new stuff with compositing (aside from some stuff that's coming later). That being said, I think shot 3 is a good example of how to approach breaking a shot down into layers for compositing, along with how you can get creative with combining things from both EEVEE and Cycles to make the most of your scene.

In this shot the green car comes round the corner and then the blue car pulls out in front (opening the parking space which will spare the former from being flattened when Rog appears in the next shot).

From that I've got the focus of the shot: the two cars. They're the things I want to spend the most time and effort on making look good. The stuff that's worth the extra time of rendering in Cycles. The rest of the parking lot and the parked cars make up the foreground for this shot. Beyond them the trees, adjacent street, traffic and neighbouring buildings make up the background, with the trees creating a nice boundary between fore and background.

Here's a simplified version of how I've broken up these layers even further, along with which render engine was used for each:

(1) Background Buildings + Street (Cycles)

All rendered as one image including the buildings and the various street props. Since nothing back here moves and neither does the camera I don't gain much by rendering all those extra frames. Rendering just one image also means I can render at a much higher quality without having to wait hours (or days) like I would with a full sequence.

(2) Background Traffic + Crowd (EEVEE)

Rendered as an image sequence. For places where elements pass behind things in the static background the objects in front are masked out of the sequence.

(3) Trees + Parking Lot Perimeter (EEVEE)

I originally included this together with the background buildings as a static layer until I decided I wanted the leaves on the trees to rustle a bit, so I rendered it as a sequence with EEVEE.

(4) Parking Lot + Parked Cars (EEVEE)

This one could've been another static layer but I decided to include the moving shadow for the green car. For the blue car I rendered its moving shadow in Cycles separately and then composited it back in (due to its close proximity to the camera).

I could've rendered both of the moving shadows in their own layer(s) and done this one as a static image layer, but the render times in EEVEE weren't too bad.

(5) + (6) Focus Cars (Cycles)

Both image sequences rendered in Cycles for better glass shaders (both the windows and lights) and more accurate reflections in the paint shaders.

---

And here's some comparisons of the same scene rendered in EEVEE (without raytracing, it's a bit much for my old GPU), Cycles and the composited version:

I still need to do some colour correction/grading to bring it in line with how sunny the overview shots are, but the final composite looks much better than the raw EEVEE render without taking the massive amount of time it would take to render everything in Cycles.

Compositing in depth of field and motion blur

Something else I've been experimenting with again is depth of field and motion blur. SFM had both, but it's implementation wasn't all that great so I rarely bothered with it.

As for Blender, EEVEE and Cycles have in-scene options for each, but since I'm compositing in things from both engines I'll need to handle it via compositing as well. Doing that requires rendering some extra passes; for depth of field you need a depth pass (how far away objects are from the camera), while for motion blur you need that and a vector pass (whether an object is moving and in what direction). You'll also need to isolate elements in the foreground and blur them separately if you want to achieve a distant focus look with depth of field.

Of the two, I'm making use of depth of field more, while I'm mainly using motion blur for smoothing out camera movement in certain shots.

What's coming up next

Next up is working on the post-stomp shots (5-7). Those footprint craters are going to add a lot of scene-building work, as will simulating the remaining bits of destruction, so things are going to slow down a bit again compared to the brisk pace of work on shot 1-3.

I'm going to focus on the most complex shot first (which is the last one) and then work backwards afterwards to finish up the rest. I'll post another update once I'm done with scene building those footprint craters and what remains of the surrounding area.