Posture Analysis App: How Phone-Based Posture Analysis Works and How to Choose One for Work

What is a posture analysis app?

A posture analysis app uses a phone or tablet camera to estimate how a person's body is positioned. Under the hood, a pose estimation model finds key body points, such as ears, shoulders, elbows, wrists, hips and knees, and the app turns them into joint angles.

What the app does with those angles depends on who it is for:

• Wellness and fitness apps compare your standing or sitting posture to an ideal, or check exercise form.
• Workplace ergonomics apps score real work tasks with an established method such as REBA or RULA, so a safety manager can rank jobs by risk and track improvements.

This post focuses on the second type. If you want to try one now, ViveLab's free AI posture analysis tool accepts a phone video or photo, runs pose detection on the device, and returns an estimated REBA or RULA score with a printable report.

This is general information, not medical advice. Posture apps do not diagnose conditions, so see a clinician if you have persistent pain.

How does a posture analysis app work?

Most apps follow the same pipeline.

1. Capture. You record a video or take a photo of the person doing the task.
2. Pose estimation. A model detects body landmarks in each frame. Google's MediaPipe Pose Landmarker, a widely used option, tracks 33 landmarks and outputs both image coordinates and estimated 3D world coordinates (Google AI Edge).
3. Angle calculation. The app computes segment angles, such as how far the trunk is bent forward or how high the upper arm is raised.
4. Scoring. A workplace app maps those angles onto the scoring tables of a method like REBA or RULA, adds any manual inputs such as load, and calculates a final score and risk level.
5. Reporting. The result is shown with the frames used, the angles and the score.

The scoring step uses the published method. The AI is doing the measuring. You can always check the arithmetic by entering the same angles into the RULA calculator or a REBA calculator.

Why does on-device processing matter in a posture check app?

Workplace video shows real people, often identifiable, sometimes in sensitive settings. Where that footage goes matters to workers, unions and legal teams.

Some apps upload video to a cloud server for analysis. Others run the model on the phone or in the browser. Google's MediaPipe, for example, provides a JavaScript Pose Landmarker that runs in a web browser, with modes for single images and for video frames (Google AI Edge, Web guide).

On-device processing means the video does not need to leave the device for the posture analysis to happen. ViveLab uses this approach. Whatever app you choose, ask:

• Is video uploaded, and if so, where is it stored and for how long?
• Can the analysis run without an upload?
• Who can see the results?
• Can workers be blurred or anonymized in reports?

How do you record video for a usable posture score?

The quality of the input decides the quality of the output. Research on video-based assessment shows that camera view and occlusion affect results. In one study comparing an OpenPose-based system with motion capture, conditions included body occlusion and frontal versus non-frontal views (Kim et al., 2021). In another, the authors suggested that side and oblique views made wrist posture hard to rate reliably (Balogh et al., 2025).

Practical recording checklist:

Do	Why
Get the whole body in frame, head to feet	Missing joints force the model to guess
Film from the side for lifting and bending, or about 45 degrees for mixed tasks	Side views show trunk and arm flexion best
Add a second angle for twisting or reaching across the body	One view cannot capture rotation and flexion equally well
Hold the phone steady at about waist to chest height, landscape	Reduces distortion and keeps the worker centered
Use even lighting and avoid strong backlight	Low contrast makes landmarks less reliable
Keep other people and equipment out of the line of sight where possible	Occlusion is a main source of error
Record at least one full work cycle, ideally two or three	Lets you see the worst and the typical postures
Note the load weight, force and frequency separately	Video cannot see these

Always tell workers what you are recording and why, and follow your local consent rules.

What can a posture analysis app not do?

Every camera-based app shares these limits.

• It sees in 2D. A single camera flattens depth, so bending toward or away from the lens is underestimated. Some models estimate 3D, but it remains an estimate.
• It misses small joints. Wrist deviation and finger posture are hard to measure from full-body video.
• It cannot weigh the load. REBA and RULA both include load or force, and the app needs you to enter it.
• It does not know the context. How often the task is done, how long the shift is, and how much recovery time workers get all come from observation and records.
• It is not a diagnosis. A high score means high exposure to risk factors, not that someone is injured.

A 2025 systematic review found markerless camera systems offered moderate-to-high accuracy and good reliability but noted small sample sizes in many studies and underreported reliability (Scataglini et al., 2025). Treat app scores as screening estimates. We look at the evidence in more depth in how accurate is video-based ergonomic assessment.

How do you choose a posture analysis app for work?

Score each candidate against these criteria:

1. Recognized methods. Does it use REBA, RULA or another published method, and follow its scoring exactly?
2. Editable inputs. Can you enter load, force, coupling and activity, and correct angles?
3. Visible reasoning. Does it show the frame and the angles behind each score?
4. Privacy. Can it process video on the device? What is stored?
5. Reports. Can you print or export a report with space for assessor review?
6. Ease of use. Can a supervisor record and run an analysis without training?
7. Cost and access. Is there a free tier or trial so you can test on your own tasks?

If you are comparing broader platforms rather than apps, our guide on how to compare ergonomic assessment software covers wearables, depth cameras and checklist tools too.

Where does a posture analysis app fit in an assessment?

OSHA describes ergonomic job analyses, observations, surveys and worker interviews as proactive ways to find ergonomic problems before they cause injuries (OSHA). A posture app speeds up the observation and analysis part. It does not replace talking to workers or reviewing injury records.

A sensible workflow:

1. Use injury logs and worker input to choose tasks.
2. Record each task following the checklist above.
3. Run the app to get an estimated score.
4. Have a trained assessor review the frames, angles and inputs.
5. Fix the highest-risk tasks and re-record to confirm improvement.

For office staff, an app is often overkill. A structured self-check and a few adjustments usually solve the main issues. Start with our guide to proper sitting posture.

Next steps

Pick one task, record it from the side for two work cycles, and run it through ViveLab's AI ergonomic assessment. Compare the score with a manual check using our RULA assessment guide, and note where the two differ.