CMAvatar3.0 Debuts! CHINGMU Teams Up with Original Force to Redefine AI Motion Capture.

In the field of digital entertainment content creation, motion capture technology has become a core productivity tool for film, gaming, and animation production.

In recent years, numerous groundbreaking works have emerged in this industry, all empowered by motion capture technology. From the domestically developed AAA title Shadow Blade Zero, enabled by CHINGMU’s motion capture technology, to A Record of a Mortal’s Journey to Immortality, China’s first animated series created using "live-action motion capture + real-time rendering" technology by Original Force, this technology serves as both an efficiency accelerator and a critical bridge connecting technology and creativity.

However, long-standing pain points in the motion capture workflow—such as time-consuming data processing (taking days to complete), unavoidable manual corrections, and missing marker data due to occlusion during capture—have shackled the efficiency of motion capture.

In May 2023, Original Force launched DeepMocap, a self-developed neural network-based motion capture solving system. Built on vast amounts of real-world motion capture data and enhanced with prior distribution-aligned data augmentation for training, DeepMocap achieved accurate skeletal animation output even with imperfect or erroneous marker data, significantly improving solving efficiency and reducing manual correction costs. Now, a year later, Original Force has refined and upgraded this technology, unveiling the revolutionary Kinetiq AI Solver, which disruptively elevates industry efficiency and addresses multiple pain points in traditional motion capture workflows!

CMAvatar Upgraded! Motion Capture No Longer Requires "Slow, Meticulous Work"

The all-new CMAvatar3.0 is officially here! CHINGMU has deepened its collaboration with Original Force by exclusively integrating Original Force’s self-developed Kinetiq AI Solver into CHINGMU’s CMAvatar motion capture system. This integration dramatically improves solving efficiency, slashes manual correction costs, resolves multiple traditional workflow bottlenecks, and elevates both user experience and productivity.

Compared to CMAvatar2.0, the Kinetiq AI Solver offers significant advantages:

1. Smarter: Handles scenarios traditional solvers struggle with—lost markers, occlusions, mislabeled positions—with ease.

2. More Efficient: Delivers fast, accurate skeletal animation data with minimal need for manual correction.

3. More Flexible: Supports 2 integration modes and 2 solving modes to adapt to diverse scenarios and requirements.

4. ...

(Comparison of results with/without Kinetiq Solver)

CHINGMU & Original Force Join Forces to Build a Cutting-Edge Motion Capture Data Pipeline

As a global leader in optical motion capture, CHINGMU, leveraging its deep technical expertise and industry experience, has partnered with Original Force to explore new paradigms for optical motion capture technology. After two years of relentless R&D, they successfully integrated the Kinetiq AI Solver into CMAvatar3.0. This upgrade not only retains high-precision motion capture capabilities but also breaks through traditional limitations, delivering unprecedented efficiency, reducing manual intervention, and setting a new industry benchmark.

The Kinetiq AI solver features two flexible integration modes, offering customized solutions for diverse motion capture scenarios.

Mode 1: Full Pipeline Solving (Marker Cloud to Skeletal Data) | Tackling Complex Scenarios

For challenging scenarios with occluded or ambiguous markers, Kinetiq AI Solver provides a fully automated workflow.

After CHINGMU’s optical cameras capture and reconstruct raw 3D marker points, artists can directly input unprocessed data into Kinetiq. The solver performs two core tasks:

First, it acts as a "smart labeler" via a neural network, automatically identifying and labeling each 3D point's corresponding body part (e.g., distinguishing between the "left wrist" and "right knee"). This replaces the time-consuming manual labeling in traditional processes with efficient AI computation. Then, based on these labeled markers, it calculates the rotation and displacement data of skeletal joints, ultimately generating smooth and natural motion animations.

The advantage of CMAvatar 3.0 in this mode lies in its almost fully automated process from data collection to motion generation. Even in extreme scenarios like martial arts with overlapping limbs or dance with high - speed spins, it can stably output usable data thanks to AI's intelligent processing.

(Kinetiq Solver accurately resolves occluded markers)

Mode 2: Enhanced Skeletal Solving | Speed and Specialization

While traditional motion capture provides high - quality 3D marker identification, the AI - based Kinetiq solver serves as an "efficient solving engine", focusing on the key task of skeletal kinematics solving.

Once the classic motion capture process pre - labels the 3D markers, artists can input the labeled data into Kinetiq. Its in - depth understanding and optimization of the labeled data allow it to process not only its own data but also markers from various sources. This enhances its versatility and operability in practical applications.

At this point, the solver can concentrate on the core skeletal kinematics solving. It can allocate computing resources to precisely calculate joint rotation and displacement data.

This division of labor, with a focus on a single process, significantly reduces the AI solver's computational load, boosting solving efficiency. It's ideal for single - person, simple motion capture or scenarios with extremely high real - time requirements.

Note: When using this approach, ensure high - accuracy marker labeling to prevent errors from affecting subsequent solving.

(Kinetiq handles complex motions and occlusions in real time)

By combining full-pipeline intelligence and precision specialization, CMAvatar3.0 with Kinetiq covers nearly all motion capture scenarios. The two modes complement each other, delivering smarter, faster, and higher-quality solutions.

Breaking Free from Tradition: Kinetiq AI Solver Redefines Industry Standards

To suit diverse practical needs, the Kinetiq solver offers two specialized modes:

Real-time solving mode can determine reasonable skeletal positions even with marker loss or mislabeling, enabling real-time effect monitoring for film crews. As the training dataset grows, the neural network continuously improves solving accuracy.

Post-processing solving mode leverages data from surrounding frames to enhance result accuracy, drastically cutting the crew's workload of frame-by-frame marker data repair for generating valid skeletal animation data.

(Skeletal Solving Comparison)

CMAvatar2.0: Errors (red) cause distorted animations.CMAvatar3.0 with Kinetiq: Flawless results, no post-processing needed.

Compared to CMAvatar 2.0, CMAvatar 3.0 with the Kinetiq AI solver better handles inaccurate or incorrect markers.

In complex or multi - person motion capture, markers often get close, overlap, or are occluded. Traditional systems without an "AI brain" fail here, producing jerky or incorrect animations that need manual fixing.

CMAvatar 3.0, backed by Kinetiq's powerful solver, can smartly generate accurate skeletal animation data. It transforms the traditional process from "system - assisted, human - dependent" to a new era of "AI - automatic solving".

In addition to reducing manual marking and calibration efforts and boosting solving speed, the Kinetiq AI solver also improves data processing, environmental adaptability, and occlusion handling, effectively addressing several pain points in traditional workflows.

In terms of efficiency, quality, and stability, the Kinetiq AI solver offers a superior option for motion - capture production needs.

[CMAvatar 2.0 VS CMAvatar 3.0 with Kinetiq AI Solver]

Data Processing: CMAvatar 2.0 requires manual data processing, with artists spending much time on calibration, marker tracking, and post-manual repair. CMAvatar 3.0, leveraging the Kinetiq AI solver's end-to-end processing, automates solving, saving labor and calibration time.

Environmental Adaptability: CMAvatar 2.0 has good lighting interference resistance but may still be affected in special scenarios. CMAvatar 3.0 with the Kinetiq AI solver shows much better lighting interference resistance and can handle complex environments well.

Occlusion Processing: In CMAvatar 2.0, occluded markers may cause data loss and need manual frame interpolation. The Kinetiq AI solver in CMAvatar 3.0 can predict occluded-part motion via a neural network, improving robustness significantly.

Note: "Robustness" reflects a system's ability to function stably despite changes in its internal structure or external environment.

(Comparison of Kinetiq Solver and Non - AI Solver Effects - Wall Collision Test)

(Comparison of Kinetiq Solver and Non - AI Solver Effects - Entanglement Test)

By overlapping the skeletal solving results, it's evident that CMAvatar 3.0 with the Kinetiq solver can accurately solve motion data even when actors are occluded or their bodies overlap, whereas CMAvatar 2.0 causes erratic actor movements, increasing manual correction costs.

CMAvatar 3.0 represents an upgrade to a fully intelligent system, thanks to the Kinetiq AI solver. It enhances efficiency in motion capture, breaks traditional application boundaries, and makes it easier for artists to capture complex motions like finger movements, quadrupeds, and multi - rigid - body props.

In addition to the Kinetiq AI solver, Yunlai Digital has other advanced technologies and is developing refinement tools via reinforcement learning for professional-level data repair. In the future, Qingtong Visual hopes to深化 cooperate with Yunlai to integrate more cutting - edge AI technologies into motion capture production, enabling smarter artistic expression for global creators.

Qingtong Visual's innovations are redefining the vitality of digital content, whether it's bringing characters to life in films, creating smooth interactions in games, or enabling realistic emotional expression for virtual humans.