PRO, PRO+, ULTRA: how to choose the right Semcam Live model

Core view: the models are not simple “low-, mid- and high-tier” options. PRO, PRO+ and ULTRA correspond to different task complexity: whether HPE, fingers or face are required; how many people and what distance are involved; whether the system will be deployed as a long-term fixed installation; and whether the project can currently accept a pre-release product. Correct selection should define the task first, then decide compute and specifications.

1. Why choosing a model by budget can lead to the wrong decision

When purchasing cameras, people are used to comparing pixels, compute and price first. But the result of a motion capture system is determined jointly by camera layout, number of people, motion speed, occlusion, real-time or post-capture requirements, output interfaces and center configuration. Higher single-camera parameters do not automatically mean higher accuracy across the whole venue; if a research project that needs HPE chooses a model that does not support HPE, this cannot be fixed later with a simple upgrade; a team doing only single-person real-time previs may not need face and large-space specifications either.

We place the three products in one system: PRO emphasizes professional real time, PRO+ adds HPE and finger capability, and ULTRA is aimed at larger spaces, more people, and finger and face requirements.[1] They are not simply low-, mid- and high-tier models, but product entry points for different levels of task complexity. At the same time, we also clearly distinguish product status: the current product FAQ still marks ULTRA as “coming soon,” so this article describes it as a pre-release product and does not use the same wording as models already available for delivery.

The first step in selection should be answering seven questions: maximum simultaneous people; farthest distance between person and camera; main motion speed; whether fingers and face are needed; whether post-capture HPE is needed; whether there are robots, tools or prop rigid bodies; and whether the data enters real-time engines, robotics, life science or animation software. Budget is a constraint, but it should not replace task definition.

Table 1 | Semcam Live model matrix (based on current public product information; ULTRA is pre-release)

Model | Current main parameters | Functional boundary | Recommended priority scenarios | Product status

Semcam Live PRO | 40 TOPS; 10W; 2.3 megapixels; up to 4 people; farthest 18m; 120fps | Does not support HPE, fingers or face | Animation previs, university teaching, single-person/small-group real-time interaction, rapid prototyping | Listed on product page; sales and delivery regions require project confirmation

Semcam Live PRO+ | 55 TOPS; 15W; 2.3 megapixels; up to 4 people; farthest 18m; 120fps | Supports HPE and fingers; does not support face | Life science, robotics demonstrations, professional animation, projects requiring post-capture precision | Listed on product page; specific configuration requires project confirmation

Semcam Live ULTRA | 100 TOPS; 18W; 5.2 megapixels; up to 12 people; farthest 45m; 120fps | Planned HPE, fingers and face | Large-space robotics data fields, LBE, simulation training, multi-person capture | Current product FAQ marks “coming soon,” so these are pre-release specifications

2. The shared system foundation across all models

The three cameras are not three separate software ecosystems, but enter the same Active Center. The camera side processes human detection, tracking, segmentation, 2D keypoints, ReID and confidence; the center side handles cross-camera matching, triangulation, 3D keypoints, body rigs, IK, filtering, retargeting, projects and export.[1][2] Therefore, selection is not only comparing cameras, but also considering Active Center hardware, network, storage and software authorization.

All three products share the same direction: multi-camera synchronization, edge AI, local processing and 120fps real time. We also provide interfaces including 250 human keypoints, PoE power, ROS, C++, Python, Matlab, MuJoCo, Isaac, OpenSim, Visual3D, C3D, Unreal, Unity, Blender and Maya.[1][2] Specific interface versions, SDK scope, center configuration and performance under different models should follow technical documentation and PoC.

The shared foundation makes progressive build-out possible: a team can start with a smaller space and fewer people, retain the Active Center workflow, then expand cameras and industry interfaces. Whether different models can be mixed on one network, how licenses are upgraded, and how historical projects remain compatible must be confirmed against the project version and solution. We will not compress an unconfirmed expansion path into a single phrase such as “seamless upgrade.”

3. PRO: designed for real-time use and accessible professional spaces

Current public specifications for PRO are 40 TOPS, 10W, 2.3 megapixels, up to 4 people, farthest 18 meters and 120fps; it does not support HPE, fingers or face.[1] Its advantage is not “the most complete functions,” but entering a multi-camera real-time workflow with lower power consumption and baseline specifications. Suitable customers are usually already beyond single-video tools, but do not yet need post-capture high precision or detailed hands.

Animation previs is a typical scenario. Directors or animators need to quickly see body rhythm, blocking and character driving, while weapons and cameras can be supplemented by Goku. Output can enter Unreal, Unity, Blender, Maya or MotionBuilder, but specific retargeting results still depend on the body rig and character. PRO has no HPE, so whether key shots require higher-quality post-processing should be evaluated before purchase.

University teaching and interactive prototypes also fit PRO. Students can learn multi-camera setup, calibration, 3D reconstruction, real-time body rigs and engine interfaces without starting from markers. For research, PRO can be used for methodology teaching, real-time demonstrations or projects with lower precision requirements, but “professional camera” should not be inferred to mean suitability for all paper measurements. If research needs HPE or fingers, PRO+ should be prioritized and comparative validation completed.

4. PRO+: the main model balancing real time and post-capture precision

Current public specifications for PRO+ are 55 TOPS, 15W, 2.3 megapixels, up to 4 people, farthest 18 meters and 120fps. It supports HPE and fingers, and does not support face.[1] Compared with PRO, the real boundary is not 15 TOPS, but the dual link of “real time + post-capture precision” and hand information. This brings it closer to the shared needs of robotics, life science and professional content production.

Robotics demonstration data usually needs the relationship between hands and tools. PRO+ with Goku can form synchronized data for markerless human body, hands and rigid-body objects. Note that our current public information says “supports fingers,” which does not mean every grasping posture can be stably parsed. Hands occupy fewer pixels in full-body views and are frequently occluded. Crossed hands, holding small tools, hands behind the back and fast operations should be tested, and any required camera positions or distances should be stated.

Life science projects care more about HPE. Our currently public information shows that HPE keypoint error can be less than 1 cm,[1][2] but this metric cannot be directly equated with joint angles, dynamics results or clinical validity. PRO+ is suitable as the main platform for joint validation: around gait, squats, sit-to-stand, running and jumping, compare with reference systems and fully record samples, camera layout, clothing, error definitions and repeatability. We will build professional credibility from validation results for specific tasks, not use a single number to cover all research questions.

5. ULTRA: a product vision for large spaces, with status clearly marked today

ULTRA is currently planned for 100 TOPS, 18W, 5.2 megapixels, up to 12 people, farthest 45 meters, with planned support for HPE, fingers and face.[1] These specifications point to three high-value scenarios: multi-person robotics data fields, large-space LBE and simulation training. Higher resolution and compute help provide conditions for long-distance and multi-person processing, but real performance still depends on lenses, coverage, lighting, center configuration and algorithm version.

Because the current product FAQ still marks it as “coming soon,” we describe ULTRA as “pre-release/planned specifications” and conduct joint testing or intent discussions with suitable partners. Early projects need further confirmation of Beta responsibilities, functional changes, delivery timing, upgrade methods and acceptance boundaries.

For ULTRA, we will not use a parameter poster alone to prove large-space capability. More valuable evidence should come from real venues close to the target scale, including floor plans, camera count, multi-person crossing paths, props, edge areas, frame rate, latency, identity recovery, long-term operation and failure clips. Face capture also needs to state distance and resolution conditions in full-body large spaces; close-up tests cannot be used as substitutes.

6. Recommended combinations for five scenario types

Animation previs and teaching: PRO + Active Center, with Goku added for weapons or cameras; key shots, high-quality motion libraries or hand requirements shift to PRO+.[7] Robotics demonstration: prioritize PRO+ + Goku, with focus on validating hands, tools, ROS and simulation; large-space multi-person data fields should be evaluated after ULTRA launch.[4] Life science: prioritize PRO+, centered on HPE, C3D/OpenSim/Visual3D and third-party comparison.[3]

LBE and culture-tourism: a small-space PoC is possible in the short term; the target solution is more likely ULTRA + Goku + Unity/Unreal, but formal delivery and validation of multi-person, distance, occlusion and long-term operation must be awaited.[5] Simulation training: choose PRO+ or future ULTRA according to space and person count, use Goku for key instruments, and implement scoring through industry rules.[6] Every combination must determine actual camera count and center configuration; this article does not infer a complete quotation from single-camera parameters.

We also state “not suitable” clearly: PRO is not suitable for projects where HPE is a formal requirement; PRO+ is not suitable for projects that need full-body large-space multi-person face and do not accept additional solutions; ULTRA is not suitable for purchases that must currently use a mature mass-production model and cannot accept pre-release risk. Clear exclusion conditions help customers judge fit faster.

7. Conclusion: the value of the product matrix is helping customers avoid paying for capabilities they do not need

PRO, PRO+ and ULTRA should be understood as three task entry points: PRO solves stable real time and professional-space entry; PRO+ adds HPE and fingers, connecting robotics, life science and high-quality production; ULTRA plans large space, multiple people and face, taking on future data fields, LBE and training. The three share Active Center, local edge architecture and industry interfaces.

For us, product status transparency is as important as the parameters themselves. Public product parameters, internal test results and third-party independent validation need to be clearly distinguished; ULTRA is still marked “coming soon.” We will not write joint-calibration accuracy as human-body accuracy, will not write HPE as real time, and will not treat interface names as unlimited compatibility. Only in this way is the product matrix not merely a sales tier, but a professional method that helps customers make the right choice.