We use Motive 2, a motion capture software platform, which helps to digitalize
a movement of a person or an animal, as well as any additional prop (a
sword, axe, broomstick, chair, etc.) in the real time workflow.
Also, you can work with virtual 3D scenes. For this goal, we use real-time translation to a third-party 3D package such as Maya or MotionBuilder. Recording is performed to two systems simultaneously: to Motive and to the selected 3D package. This feature allows an actor to go round 3D objects of a virtual scene or dodge other characters, for example, an attacking dragon that is present only in a 3D scene. This allows you to check immediately, right during filming, whether the actor was oriented correctly, whether the actor has moved enough and whether there was any overlapping with other virtual animated characters.
Using NatNet makes it possible to integrate into Unreal Engine 4, so that the actor — while being displayed in the 3D scene by Unreal Engine — can at the same time see everything around through the VR system (currently optional). It definitely helps to achieve the maximum effect of presence in the 3D scene.
The working capture area is a 17x21 meters rectangle, with a height of more than 5 meters. 48 high-resolution cameras are located at a distance of 2 meters from the working area. Up to six actors can simultaneously participate in the shooting. To shoot an actor falling on the back, a special control marker is placed on an actor’s chest. The floor has a special rubber coating that helps to break the fall, plus you can use additional soft mats.
Diagonally, actors can run up to 30 meters. Technical entry gates are 3.5 meters high, which allows you to bring large-scale props, pieces of scenery and large animals.
The standard recommended capture rate is 120 frames per second; FPS change is possible when data is exported. The upper limit to the capture rate is 360 frames per second — for objects moving faster than 100 meters per second.
Motive software used at our studio helps to avoid grueling cleaning of animation due to the following built-in features:
Recorded marker data are processed and exported to files of the preferred format:
C3D — contains a point (marker) cloud.
FBX_Binary — for bones.
FBX_ASCII — for Skeleton dolls.
BVH — for Skeleton dolls described before the shoot.
CSV — for optical and rigid body data.
Our studio can prepare any additional props for the shoot by placing markers on them. However, we do not recommend using rare/antique items, since placement of markers involves using screws, tape or stickers that can damage the valuables.
To achieve the maximum result in an optimum manner, it makes sense to create a framed model of a prop in order to reduce the area of markers overlapping. For example, the best way to shoot a knight's shield is to make its copy out of metal or plastic wire frame mesh, so that the shield is as transparent to infrared light as possible. The more visibility through the shield you can reach, the better the result will be. We do not recommended to make a shield out of solid transparent materials (glass, organic glass), since many of these materials are not transparent to infrared light and can reflect it, thereby worsening the results of shooting.
If any one-of-a-kind props are used, they should not have reflective, glossy or glazed surfaces; otherwise, these surfaces should be isolated by light-absorbing material. The use of synthetic materials that intensively reflect infrared light can also worsen the results of filming. For more realistic perception during filming, the real prop model can be displayed in the View window by using a 3D copy of the prop prepared in the OBJ format (with a minimum number of polygons in order to maintain the system's performance).