Initial Concept

The first step of our project was a concept and paper prototype. The project we had in my mind utilizes just the volumetric pass, a pass that captures just the light rays over flat, black shaded assets so that the light rays can be added and adjust in composite rather than be integrated with the beauty. I’ve always thought these passes were very beautiful.

Prototype

I created quick concept art renders in Arnold. I started using Mixamo to search for preexisting dance animations. Mixamo is a site, now available through an Adobe account, where there are existing rigs and characters and animations that you can use to quickly prototype something. Most of the animations were more modern so I instead picked three different poses. I exported the FBX files into Maya and used Arnold’s aiAtmosphereVolume. You can find this setting under Render Settings > Arnold Renderer > Environment > aiAtmosphereVolume and then going to the Attribute editor tab.

I usually start by setting the density to 0.050. I then put an aiUtility Shader onto the figure geometry and set the shade mode to Flat. This makes the shader have no diffuse and no specularity so the object is just black.

Then I added one spotlight and one skydome with really low intensity to the scene. I positioned the spotlight and camera on opposite sites of the figure till I found positions that showed off the volumetric light rays best. I think I would go more creamy with the color but I’m happy with how the shots turned out.

The next stage was to create a presentation detailing our project scope. This presentation also would serve to explain our project when we cast for a dancer. Below are the presentation slides.

We also created a production timeline to make sure we were on schedule for the project.

Here is the first draft of the email we sent to the Dance department in search of prospective collaborators.

We experimented in Unreal prior to our motion capture sessions. First we removed all the lights from the scene then added a soft point light. From there we added fog to the scene and then figured out how to create unlit shaders (similar to a flat shader in Arnold). We struggled to get the cast shadows from the shapes once adding a spotlight. Here is a video of our progress with adding fog and trying to get the cast shadows. At the end of class we did not succeed but we made good visual progress in the right direction.

Once we heard back from the dancers a motion capture session was scheduled. We recorded and edited the process working inside the motion lab.

With the fbx files from our motion capture session I attempted to just import them into Maya to see what would happen. The rig was not where it was supposed to be and everything was a bit of a mess. This process clearly involved more research. I checked in with Vita who recommended I watch the videos she shared with us about the process. I watched several of the videos and found the terms I needed to search in order to find the right tutorial for the task I needed.

The important lessons I got from the tutorial were to import the motion capture session fbx into the scene and select FBX Merge > and the name of your take. Takes are what you name your motion capture recording data.

As I watched the tutorial I saw the creator was using a Mixamo rig. Mixamo is an easy autorigging web application available to us at OSU through our Adobe Creative Suite.

I had a model I’ve been working on in my side time and I placed that model into Mixamo and followed the easy steps to get a basic rig. The model I had was in A pose, not T pose, but Mixamo had an option to export it out into T pose. T pose is important to make things easier in MotionBuilder. BUT the Mixamo Export, even with the T pose checked, came into MotionBuilder in A pose. I decided to check the same Mixamo Export in Maya and it did come in as T pose. So I added an extra step of Maya into the initial rapid rigging pipeline of Model >Mixamo > Maya > MotionBuilder to get a proper T pose model.

For the rigged character, when you import it into the scene, you select FBX Merge > <No Animation>

Once in the scene you don’t scale your rigged fbx character but you scale and adjust your motion capture data instead.

Another important step is called Characterization. In the asset browser you assign Characters > Character to the hip of your rig and the motion capture rig hip. You define which Biped for a human and then label one as your Target and one as your Motion Capture data. The definition step allows you to check if your rig lines up with the expected rigs. Using a Mixamo rig meant it lined up immediately for me.

The next import step I needed to learn was to target the Target character and have it Source to your Motion Capture Data character. After that step the rest of your time in motion builder is spent refining the connection between the two rigs. Right now, the two rigs don’t align perfectly and there are some awkward bends in the character joints. But, for now I wanted to get a rough targeted motion capture out into Maya to test with the prototype lighting setup. There was an easy button right in the File dropdown menu to export it live to Maya. Once in Maya I threw on the flat black shader, added aiAtmosphericVolume, and then a spotlight. Everything worked really well. I found some interesting poses and started exploring lighting setups and camera angles.

At the start of spring break I created a full motion test in Maya with the intention of using interesting shots I was able to capture for our Unreal edit. While we planned to do our final edit in Unreal, we ended up running out of time so my back up tests ended up becoming our final submission in the video at the top of this page. I did another pass using a different camera and did edits between the two for the final submission. Below is the first pass in a single camera