The Director of the upcoming horror film Relicts Arkadiy Demchenko has told us about the production process behind the film, discussed character rigging, mocap, and hair creation, and talked about the hardships the team has to go through to deliver a fantastic result.
In case you missed it
You may find these articles interesting
My name is Arkadiy, but usually, we go with just Ark, because most non-Russian people can't spell it without a typo. I have 15+ years of experience in CG, mostly for film & commercials. My interest in CG started with digital humans and facial rigging specifically, but since we didn't have much of a demand for that in Moscow, I'd slid into other fields – lighting and compositing – where I stay professionally these days. So, it wasn't my original interest, but I've loved it when I started doing it.
I did little stop-motion animations and graphic projects since school, so, the idea to make a 3D animated film has always been out there. It had transformed into action somewhere in 2011 while listening to the music by two profoundly eerie composers – New Risen Throne and Lustmord.
The former one has given birth to a bleak atmosphere of an isolated settlement shackled by snow and ice in my brain. Then a certain track by the latter one had sparked a climax scene of a dark and bone-chilling story in this environment. Influence from H.P. Lovecraft stories and the Silent Hill game series had crept into this mix, and thus "Relicts" was conceived.
Generally, that's my way of generating ideas – daydreaming while listening to music. When it resonates, visions start to get born. I'm happy to say that Gabriele Panci, aka New Risen Throne, is the main composer on Relicts now. I have a rare opportunity to work with the artist who has inspired the original idea.
The development started when my colleague, Valeriy Zrazhevsky, had painted the first set of concept art for our main character. She never got a name and is still known as just the "Girl".
Shortly after, Alexander Ovchinnikov had developed the second character, her husband. Not to ruin the creative style, we called him the "Man".
A bit later, Anna Mitura-Laskowska joined the concept process and fleshed them both out with a series of moody portraits and scenes.
Meanwhile, several artists contributed their vision of our gloomy environments: Tom Honz, John Park, Andreas Rocha, Piotr Jablonsky, Reza Afshar, Igor Vitkovskiy, Gregory Lozinsky, and others.
The project was boiling in this concept stage for several years until we finally managed to sculpt the characters in 3D – Marcus Fall ended up doing both of them. It was the time of raytracing coming into fashion with Mental Ray and lightmap-based SSS that took 30 minutes just to kick the first bucket (mostly due to the amount of hair and fur we would stick in it), so, it was quite a struggle until I'd jumped to Arnold renderer and was able to render everything without burning my brains out.
Right now our pipeline is settled on Maya, Arnold, Yeti, Nuke, and DaVinci Resolve.
The key idea I've learned from the start is that I can't do everything by myself – I have to rely on other artists, specialists in their fields. That's probably the most exciting and most devastating part of the production – collaborating with the right people who raise the project to a whole new level... or wasting time on the wrong ones. So, some of the tasks were accomplished only by the 5th or 6th artist who tried, which sounds a bit crazy, but the result was totally worth it.
The hardest technical thing to do was facial rigging. We've struggled for a couple of years with that and, eventually, Dario Triglia had finished body rigging for our characters and took the matter into his hands. He has built a series of custom tools and a setup based on facial scans we did with the help of Andrey Krovopuskov and his team at R3DS.
Since the film is visually realistic, mocap was a logical thing to do. Dario has prepared a mocap version of the character and a tool to transfer its data to our final character rig via Maya HumanIK. We tried OptiTrack camera-based approach in a studio called "Pilot" and Xsens costume in CGF. The former one gave us more details and grit but was ridden with artifacts. Anyway, both required a considerable amount of manual keyframe work (motion from the actress doesn’t always stick perfectly to the character due to proportional differences, and we had no fingers or facial data, so, all that had to be done completely by hand). As the result, mocap data has rarely ended up in the shot “as is”, without the cleanup and manual adjustments.
In terms of pipeline, our character consists of several "versions" – one simplified for receiving mocap data, another one with complete deformation rig for keyframe animation, then the geometry cache data flows into dynamics version, and, eventually, the simulated result goes into lighting/rendering version. Each version can have several parts loaded as references as separate scenes (e.g. hair or fur setup).
This modular approach is taken to provide each department with clean input data without unnecessary "baggage" and to allow different artists to work on their tasks in parallel. For example, Vadim Zhiltsov, our grooming TD, creates fur for the coat and submits his scene (containing just a mesh for the coat and his fur setup), and this scene is referenced and connected to the character in a master scene via blendshape. If he needs to fix or update something - he makes changes in his scene and it propagates into the whole production. The same happens to faceRig and bodyRig – they are separate scenes referenced into the final character asset and connected there.
We've paid special attention to the eyes. Besides solid modeling, the key thing is to add a "fleshy" effect – the eyelids and skin around them moving with the eyeball, and cornea bulging and sliding underneath. We've added automatic eyelids movement as well – when eyes look down, eyelids follow as well to maintain the natural position (animating this manually would be a total pain in the back). It's also very important to make a surface for a liquid between the eyeball and eyelids – it catches the highlight and makes the eyes wet and alive. This surface has to follow the curvature of both the eyeball and the eyelids to avoid major penetrations that can ruin shading.
To help the dramatic effect of the eyes, I've made a tool in Nuke to adjust the iris and areas around it – it's based on two custom masks coming from rendering and allows to put an accent in comp without a hassle of catching that fleeting position of a light in 3D.
Cloth and Hair
Cloth and hair dynamics are done entirely with Nucleus in Maya with intricate nCloth and nHair setups. Each piece of cloth is driven by simplified segments colliding with each other. The setup involves several steps of splitting and wrapping and merging again, so, even though I've made the entire thing, I get a bit scared every time I have to delve into it and change something. But now I pose the character, turn on the switch and it works from there – brings tears of joy to my eyes.
It's not very fast, but quite reliable unless the character's body (that acts as a rigid collider) penetrates itself and squeezes the cloth in-between – an elbow dips into the ribs for just a frame and the whole thing explodes! Hours of simulation lost! It sounds quite sad, but usually, it feels funny because of the way it looks – our Girl starts to walk completely naked with all her clothes wrapped around her head.
Her hair is made with nHair and uses its dynamics. It's split into sections – braid (driven by joint chain that can be controlled manually or dangle on a dynamic curve), locks of hair (driven by Nucleus dynamics), and a static volume of hair, a "hat". The main trick was to make all that work together with and without simulation – we can pose the character with hair intact and then simulate from this new starting pose just by turning a switch on.
My best advice for making a short film, besides not doing it, would be to use your strengths and find the right people for the rest. It will be hard, slow, and frustrating, many artists won't even reply, some will take the task and never move a muscle substituting promises for results, but there will be a few gems that will help you tremendously. I'm myself a Lighting & Compositing Artist, I can't paint concept art, model believable characters, or rig their faces, so, being alone I'd probably be showing nicely lit spheres and cubes right now. But I was lucky to meet and inspire some amazing artists.
One important thing to keep in mind – everything costs money. Enthusiasm fades in most of us when we inspect a monthly bill that has to be paid somehow. You can meet skilled and experienced people who offer to work for free, but these are rare exceptions that happen only when you're at some advanced stage of production, have something appealing to show already, and working on your project brings some other benefits. So, don't plan your project relying on free help, and better start saving your earnings to feed this ravenous enterprise on your own.
But even better advice would be this – don't believe anything I say and try for yourself. The journey is what really matters most of the time, right?
Our weak spot is character animation – it requires a ton of time and patience, skill and experience, back and forth adjustments and reworks (we've found that facial expressions can look quite different when rendered with lighting and had to render-test every animated version), especially so for realistic characters. That's why you see a lot of stills and not that many animation sequences.
We have strong animation artists eager to jump on the task, but we need a budget to make them forget about those pesky bills piling up and dedicate all their time to "Relicts". Therefore we've launched a Kickstarter campaign to raise it. If you can help us reach the goal, the characters will come alive and tell you their dark and terrifying story, and you'll have unique items in return for your support, so, I'd like to invite you to check our presentation and learn more about the project from the mouth of its main character:
You may find these articles interesting