FXWARS! TESLA - Trey Harrell (Tribute)

#1

So, I think that I’m going to use this as an excuse to re-study a classic film that’s among my favorites: Metropolis by Fritz Lang (1927). Many of the effects sequences in the film use adaptations of Tesla devices – in fact, we’ve got Tesla coils, logic gates, automatons, vacuum tubes, reciprocating motors, airships and all sorts of other stuff going on in various shots. Some of them depicted realistically, some of which depicted in a mad scientist sci fi light.

The film itself is the origin of much of our modern filmic language when we’re dealing with science fiction or dystopian ideas. 3 guesses where the design for C-3P0 was most stongly influenced :wink:

Because I’m almost definitely going to be working in black & white, I think I’m going to have enough time to try my hand at intercutting 2-3 effects sequences like a trailer, among which could be:

  1. an establishing shot of the city/new tower of babel (airships, electrical trains – admittedly a bit of a stretch for this challenge, but hugely iconic for the film itself and it just looks like fun if time allows)

  2. a sequence with a worker in the city operating the babel elevator computer (in the film, humans are a physical bridge in the city/machine logic gates – the theme that the workers are enslaved to the machinery, but inseparable is dominant in the film. This is a strict reversal of how logic gates – ie circuits – generally work. Here, the computer is telling the operator what to do in a quite literal sense.)

  3. the transferral of Maria’s visage and psyche to the robot (called alternately Hel, or False Maria).

#3 and #2 have the strongest Tesla connections, with #3 definitely the strongest connection of the bunch. I’m going to prioritize these 3, 2, 1 according to how quickly the work goes.

The thinking is that the usual stuff I spend tons of time on (shader/look dev, music, color) will take a backseat to the sequences and effects themselves because I plan on mimicking the film’s look which should free up time for more sequences.

Everything will be CG – some combination of 3D/Compositing/Matte work. I won’t be using any live footage for this.

As the film itself has been in the public domain for ages, I’m not going to worry if I match sequences too closely or lift music from the score. In other cases, I’d art direct something of my own “in the style of”, but the aesthetic is so iconic in this that I’d like to try my hand at nailing it – quoting directly from the film itself makes sense here, I think.

Reference shots of the stuff I’m thinking about annotated below:

The city of Metropolis hero shot – airships, electrical trains and archetypal art deco coolness all around.

A logic gate (circuit) operating a human, instead of vice versa.

Maria on Dr. Rotwang’s table. We’ve got coils, transformers and capacitors aplenty here.

Hel being transformed into a false Maria. Lots of energy discharge, and note the pentagram which could either symbolize a golem, or satanic influence (probably the latter – the film makes a strong old testament parallel with the story of Babel, as well as character names). Interestingly, the scientist (Rotwang) fuses many fringe alchemical ideas with his electrical devices (which can be seen in later shots).

Longer shot of Rotwang, Maria and False Maria, as well as an enormous Tesla coil on the ceiling (framed too perfectly in the pentagram to be a coincidence) which is a catalyst for…

Transformation! (or violation of Maria’s virginal form and psyche, if you want to really dissect it)

Gadgets in Rotwang’s lab and on the table closeup. Vacuum tubes, lights and transformers.

More gadgets, this time with what look like stills, some capacitors and more vacuum tubes.

More tubes, this time carrying what looks to be plasma of some sort.

#2

So, I sat down and made some notes on the film while I was gathering reference.

Then I figured what with this being all about the Tesla gadgets, I thought it best to start early development work on how I’m going to achieve the various necessary machinery/electrical FX for the shots I’m thinking about.

So… let’s start at the lightning!

First off, I don’t want to rely on off-the-shelf plug-in effects for anything, and if I use them, they’ll be there to augment my hero effects. There’s a ton of plugs that do lightning, and do it well (thinking of After Effects/Tinderbox off the top of my head), but I want to be able to move the camera in 3D space without manually tracking stuff, so at the very least I’ll need 3D trackers to guide my 2D effects in post if I decide to go the 2D route. If I decide to go the 3D route, I’ll need trackers to direct the electricity.

It’s looking like a 3D FX rig with locator points that can bake and export to a compositing package is the obvious starting point for either scenario.

I’ve put together two really fast concepts about how to create and be able to direct lightning arcs in 3D space, and get locators in 3D space into whatever compositing package I end up using (haven’t decided on Shake/After Effects/Motion yet… might be some combination of all of 'em for precomps because I’m going for efficiency).

Nothing special on these, just R&D playblasts at the moment as proof of concept before I start working on more complex rigs.

The first is using pimped out Maya hair dynamics based off of circular arc control curves (noise expression drivers in addition to physics drivers). I’ve got everything clustered to control locators and the sphere stand-in for my Tesla coil. I’m testing to see what kind of motion I get out of the system, and how flexible the approach is for art direction and animation purposes.

Lightning Test 1

The second is using a (very quickly) hacked version of a Maya script which uses the same approach as the maya electric arc generator, except it does curve (instead of nurbs) output, and an arbitrary number of control points.

Lightning Test 2

Both approaches are pretty directable for my hero forks, but I’m leaning toward the hair system approach as it gives a little more organic/interesting movement to my eye, although the hair system test looks less like lightning than the stock script… I’m certain I’ll get better results in the long run.

I think my next test, I’ll drive my arcs with overlapping hair curves. One for low frequency (overall “big curve” curvature) and one for high frequency (“short jaggies”), then blend between them for the final output. I’ll probably rig clusters parented to locators on my low frequency curve for direction/animation.

From there, I’ll probably rig up some sort of noise function on my end locators and add some ghosted geometry, then plug collision into my ground plane so I can get some particles spawning randomly when the lightning really connects with a surface. After comparing older rotoed lightning and newer stuff like The Prestige, I found that the ground contact bloom really sells the effect and anchors it to a physical reality you just didn’t see with handpainted roto work… and that’s something that’d be sorely missing in a 2D post-only approach unless I manually added all those ground hits. Eww.

Once I’m comfy with an approach for the electricity, I’ll make a stock electric arc rig that I can reference in my layout scenes, then lock down other simpler effects like plasma, steam, vacuum tubes, and the transformation effect.

–T

#3

Ok, I settled on a hybrid 2D/3D approach to my lightning and just ran a quick render to make sure it’s the direction I want to go.

It’s dynamic hair curves in Maya which are lit up like neon tubes casting GI lighting all over the scene (with a time-based noise mapped to their incandescence), then given particle flow over their length to give a jagged look. Then the curves are augmented with After Effects post lightning to get the small forks going, and I’m adding a few extra particle spark effects here and there.

Pretty happy with the approach and overall look now, and I think it’s time to move on to the vacuum tubes.

Animation Test:

http://www.treyharrell.com/fxwars/wip/lightning-test3.mov

Still:

#4

Very nice!

I am also testing an hybrid 2d/3d approach but it doesn’t look as good yet.

Maybe a bit too bright, but the concept works well.

#5

This is looking very good Trey.
I like your idea of working in black and white.

#6

Thanks for the replies guys!

Ulb: yeah, it’s a bit on the bright and over-the-top side right now. I’ll be toning that down a touch in my final rig, although I’m likely to keep it slightly sci-fi looking on purpose. The effect will certainly be more pronounced than in the scene I’m quoting, though.

On this one, I doubled up on the bloom during the render and black/white diffusion in post to match the look of the film stock. I might have gone a little overboard on the diffusion, but there’s no point tuning it further until I’ve got some stuff in the scenes with contrast.

Something interesting… the film in its present form has been assembled from surviving footage from all over the world: negatives, release prints etc, although almost all of the shots have heavy vignetting and diffusion (likely from the erosion of the film stock). It’s a really unique look.

An interesting note is that in the captures from the film, there’s no true white (or black) in anything… the dynamic range is super compressed with the midtones really cranked (and the top and bottom are extremely diffuse & scattered).

–T

#7

Good job man,
I don’t know maya so i wonder how are you using these hair curves:)
can’t wait to see it finished:)
cheers

#8

In Maya, you can convert any kind of curve into a “hair” curve… which just means that it’s a curve that’s affected by physics and fields. You can do ropes, tails, cables, antennas, and all kinds of cool stuff with the hair system, and it’s very fast to work with.

I set each strand to lock their start and end points to my locators, and rig clusters to groups of curve points so I can animate the silhouettes on top of heavy dynamic turbulence. In that last test, I used more noise functions to animate the clusters for me, and then I blend between the physics output and my guide curves with animation.

After that, I extrude a tube along each curve and make it heavily incandescent, and I flow a ton of particles over the curves with lots of noise, clumpiness and varying size to make the curves nice & jagged at a higher frequency than the curve’s overall turbulence. This is rendered on a layer by itself for heavy post doctoring.

I also emit a bunch of spherical area lights from the curve, set to live for 2 frames. The area lights have their intensity set to a noise function like:

areaLightShape1.intensity= 1 * noise(time * 4) * 12 + 10

which gives me lots of long-range flickering on the ground, which I don’t get too convincingly from incandescense alone on the curves, and keeps me from having to manually animate everything flickering (it’d be a TON of keyframing to do it by hand). That same type of function is used all over the place to animate practically everything procedurally at this point (including my rigged locators), and the animation can be baked out if I need to do something like export my locators to after effects for tracking.

This is rendered on the background plate with all of my scene geometry, along with the reflections of the extruded incandescent tubes (in a reflection pass).

In compositing, I then use all my locators from Maya as a guide to place secondary lightning effects (the little forks, etc) on top of the big hero arcs that are already rendered out.

For me, it was important to have something physically casting light and at least heroi arcs reflecting in my 3D scene, or I might have just done it all in post.

Once I’m close to done with the actual shots, I’ll post up some captures of my rigs and compositing breakdowns.

–T

#9

So, I took most of the weekend off, but I did manage to focus my scope to something that’s more manageable in the timeframe. I think I’ll just be doing the lab sequence, and a pretty heavily condensed version of it (the sequence in the original is almost 8 minutes long!).

I'll be working with four intercut angles:

1. A master shot of the table and robot in the chair.

2. A closeup of maria on the table

3. A tracking from medium to extreme closeup of the robot during transformation. I'll likely be heavily reworking the transformation effects from the original. Maybe animated displacements? Dunno yet.

4. The scientist at his controls, tight medium with shallow DOF.

I keep having to remind myself this isn't a modeling challenge :) I could spend a month making and shading all of the odds and ends in the lab, but I think if I just limit the scientist to behind his table I won't have to worry about cloth sims or spending too terrible much time rigging him. I can also cheat an awful lot of detail in texture seeing as how I'm working in black & white at a medium shot.

I’m skipping a reverse angle of the scientist from the robot’s POV over the table. Lang crossed the line of action an awful lot in this sequence which is really jarring now that we know better than to do that sort of thing. I could use Maria’s face as a “handoff” transition between the reverse angles, but I’m just wasting render time if I do that.

I think that I'll finally be ponying up for CrazyBump (auto normal/spec/diffuse/occlusion map generation from photo textures). Been looking for an excuse to snag it, and I'm certain it'll be a massive timesaver for this:

[http://www.crazybump.com/](http://www.crazybump.com/)

--T
#10

Hey man, that’s looking pretty sweet, great last test animation you got. Calculating the relections for the lighitng will be tricky on all the reflective materials…not sure the best way to tackle that either, I might cheat on my scene and cover most of the windos, haha.

Anwyay, good luck, keep the upadtes coming.

#11

Here’s a still test frame I did on a very quickly modified Zbrush base mesh (poly smooth added in Maya… you can see facets in her lips currently).

There’s no bump, displacement, or anything other than diffuse and (very very rough) specular maps on her face at the moment influencing SSS. The hair is a 2D billboard cheat, which I’ll likely be able to get away with because the camera parallax isn’t going to change.

Basically, I’m seeing how little I can get away with blendshape, rigging, texture and detail-wise on my characters… because of the style and selective cuts / camera angles, I don’t think she needs anything much more than more eye makeup and a proper displacement map for her head and neck.

I’m rendering out a mix test right now with three very rough blendshapes (eyes x-wide, closed, hero sneer) to see if I can avoid spending the time setting up a proper Osipa-style facial rig with hundreds of lovingly crafted muscle contractions I’ll never use. The basemesh I used for this really isn’t set up for decent animation, so I’m going to avoid a full facial rig if I can. Efficiency is the goal here, and I think she’s pretty close to good enough now.

More once the sequence finishes rendering,

–T

#12

Here’s a first stab at shaders and expressions on the Maria model.

I think she’s pretty close to working already, eye and lip weirdness notwithstanding… but I’ll be painting a proper displacement map and doing a little more eye makeup for the final. Might also wrap deform the eyebrows on top of the mesh because they’re too soft for my tastes.

Eyes open, eyes closed and a slow sneer are all I need her to do :slight_smile:

http://www.treyharrell.com/fxwars/wip/mariatest1.mov

Now onto the robot and scientist…

–T

#13

Its animated, but i have to let the computer sit over night or something to show people. i used particles so its really frustrating. but really gives the effect i was trying to achieve. I tryed using splines and noise but it was to hard to get the erratically look of lightning.

#14

Ok, did my final sculpting, rigging & texturing on the Maria character and I’m going to call her done for now. I could keep tweaking her for another couple weeks, but I think she’s pretty good now… besides, I’ve got a lot more modeling to do!

A few expression stills, and a raw render before major post effects and hair below (just diffuse bloom). I think she turned out pretty well, and is a pretty good likeness of the actress given the timeframe.

–T

#15

Well, I spent the entire day tuning to get my renders down under 10 mins a frame on my quadcore… juggling filter size/bsp/sampling/final gather etc etc etc. As any mental ray user knows, half the time you spend more time getting a flicker-free look that’ll render in your lifetime if you decide to do almost everything in-camera (which I did due to refraction and bokeh depth of field on everything in the shot).

Anyhow, I finally dialed in the first two shots and fired them off to render for a couple days (doing beauty/occlusion passes plus an RGB split alpha pass because I’m not separating layers).

So… now that every machine in the house is tied up for a while, it’s time to play with some looks! :bounce:

I know originally I’d planned on working in black & white and try to emulate the film’s look exactly… but I’ve been working in color up until post and I’ve come up with some pretty nice looks… so, survey time.

What do you guys think? Which look is your favorite?

–T

#16

wow this is really awesome…you matched the fritz lang look very close!

#17

Great work , I like 2 and 4 look the most:)

#18

Look no.4 is my choice :slight_smile:

#19

Thanks for the feedback guys! Still having a hard time deciding, so keep the suggestions coming! :thumbsup:

Here’s a second-ish of each of them in motion back-to-back… had to make sure that the final gather flicker and bokeh grain didn’t waste render time, and that the added stock grain matched the treatments.

http://www.treyharrell.com/fxwars/wip/tablelooks.mov

Worth noting that my bottled plasma widget is (you guessed it) Maya hair curves with turbulence driving extruded tubes that have a fast perlin noise driving opacity and incandescence, then passed to shader glow.

Because the shader glow is a post effect in Maya, I’ll be rendering a light flicker pass after the main plates are done. I’m a little worried that her neck and camera-side of her head are looking flat, so I might throw in a touch-up shadow pass at the same time.

And… now time for me to wait a day or two for these two cuts to finish rendering.

Edit:

I figured I should probably mention something about Mental Ray bokeh and render times.

The bokeh lens shader gives really awesome depth of field and lens defocus bloom dot effects, but it’s incredibly expensive to sample high enough to remove all of the graininess.

In fact, anything above 4 samples in my scene shot my render times through the roof, so I decided to uh… fix it in post and stop wasting my time trying to tune it out with 10 minute test renders (10 minutes if I was lucky, that is).

Anyhow, in after effects, I use:

Effect->Noise & Grain->Remove Grain (tweaked to taste per shot)

Effect->Noise & Grain-> Gaussian Blur (to make the remaining bokeh grain soften and expand a little look like film grain, by about .2 pixels)

Effect->Noise & Grain ->Add Grain (setting dependent on stock look, but I never go much higher than 30-40%)

You’ve got to be very careful that when you add grain in post that it doesn’t call attention to itself. The default settings in after effects are WAY over the top.

Hope this helps someone deal with bokeh or even final gather noise. Quite often it’s way faster and easier to just deal with picky, time consuming issues like that in compositing.

#20

I’d like to see a mix between the second and third images. I like the skin on the second image but I like the environment better on the third! Awesome stuff!