A desktop PC with a CUDA-capable NVIDIA GPU (graphics card) that has at least 8 GB of memory and compute capability 7.0 or higher, plus at least 24 GB of RAM. Theia3D can use more than one GPU. Theia recommends 32 GB of RAM, an eight-core processor, and a dedicated desktop rather than a laptop, since laptops process more slowly.

Theia3D needs at least 6 Miqus video cameras, and Theia recommends 8 for most capture volumes. One QTM computer runs up to 16 or 24 video cameras, depending on its processor, and larger labs link several synchronized QTM systems. Every joint must be visible to at least three cameras.

Yes. Qualisys video cameras calibrate together with your marker cameras in QTM, so a single capture can record both. Qualisys calls this combination True Hybrid. A Miqus Hybrid camera records either markers or video in a given capture, and you switch modes from a button in QTM. Theia3D tracks only people, so marker cameras in the same capture can track equipment or objects.

The four steps above show each stage. Theia3D processes video after capture, not in real time. Processing time depends on trial length, frame rate, the number of cameras and people, and your computer. For whole sessions, Theia3D Batch works through a list of trials without supervision. Theia recommends checking the calibration and a sample trial first.

You calibrate in QTM with the Qualisys wand and L-frame, and Theia3D uses that calibration directly. Miqus video cameras switch to marker mode automatically during calibration, which lets them detect reflective markers on the wand. Qualisys calibrates each camera's lens in-house before it ships, so your lab doesn't need a separate lens calibration step.

QTM also synchronizes inertial sensors, eye trackers, and instrumented treadmills, and Theia3D supports setups with Bertec, AMTI, and TreadMetrix treadmills. Because Theia3D uses the QTM calibration, its results share the same global reference frame as your force plates, with no separate alignment step.

Theia's packages include Visual3D. When you open a Theia3D file, Visual3D applies a ready-made model scaled to that participant, so there's no model to build by hand. Qualisys also offers analysis modules for gait, running, and baseball that take a session from capture to report in a few clicks. Labs that write their own analysis can work with Theia3D's exports in Python, MATLAB, R, or Excel.

Theia3D triangulates the keypoints from every camera view into 3D, using the camera positions from the QTM calibration. It then scales a skeletal model to each participant and fits it to those points, frame by frame. The output is the position and orientation of each body segment, and joint angles come from those segment poses. Visual3D can also calculate joint powers from force plate data. Theia publishes the full model definition in its documentation.

Your lab does. You decide where the video, pose data, and reports are stored, and who can see them. You can share results with collaborators while keeping to your privacy and data-sharing requirements.

Theia3D is backed by 60+ peer-reviewed studies, with published studies on Qualisys hardware covering gait, sprinting, pitching, throwing, boxing, and balance, in adults and in children. Many studies compared Theia3D with Qualisys marker cameras recorded in the same QTM session. Agreement varies by task, joint, and plane, so check the published results for the movements you study.

Theia3D processes video after capture, not in real time. Each joint must be visible to at least three cameras. Rotations about a segment's long axis, such as hip and knee rotation, are less accurate than flexion and extension, and a 2025 systematic review also found poor accuracy for hip flexion. When people overlap for long periods, tracking accuracy drops. Fast movements such as pitching and sprinting need high frame rates and short exposures to avoid motion blur.

[Kyle to answer: Theia, Qualisys or both, and who handles support.]