Solutions · Fitness and wellness
For fitness and wellness brands: a lab behind your movement features
Movelytics works with fitness and wellness companies that want their app to understand what a body is doing. We bring models that run on your users' phones in real time, and a team that has shipped them. You keep the product, the brand and the data.
Last updated:
Push-up · elbow angle and body lineIllustration. Authored pose sequence, not live inference.
01Your problem
Video libraries do not know what the user did
A workout video plays whether the user moves or not. It cannot count, correct or adapt. The apps that keep people are the ones that respond to the person in front of the camera.
Building that in-house means hiring for computer vision and mobile performance, then maintaining a camera and model pipeline on iOS, Android and the web. For most product teams it is a detour of many months away from what makes their brand different.
02What becomes possible
What the technology makes possible
- Counting that the user trusts
- Repetitions detected from the movement itself, in real time, on the phone.
- Feedback during the movement
- Joint angles and a form score while the exercise is happening, not in a report afterwards.
- Tests and progress
- Jump height, flexibility measures, comparison with a reference movement: numbers a user can follow from week to week.
- Every phone, not only flagships
- We optimise for the devices your users own, including low-end and five-year-old Android phones.
How it runs, step by step, is on the technology page.
03Which route
Product route or custom route
01
Start with PoseTracker
The SDK covers common exercises and measures. Your developers integrate it; there is a free tier to test on real phones. Why we built it.
02
Go further with the model
The PoseTracker Model, in private beta, adds feet, trunk and spine points and targets older phones.
03
Build with the lab
A movement of your own, a keypoint of your own, a scoring method of your own: the lab builds it with you.
If your question is how pose estimation fits into a fitness app in general, the PoseTracker guide on pose estimation in fitness apps is the developer's reference.
04Privacy
Your users' bodies are not our dataset
A fitness app films people at home, often in sportswear. That video should not travel. With on-device inference it does not: the frame is analysed on the phone, and Movelytics receives a usage count for billing, nothing else.
Your app keeps the results it chooses to keep, under your own privacy policy. We do not build a movement dataset from your users, and we never sell data. The commitments are written out on the data commitments page.
05FAQ
Questions from fitness and wellness teams
Can Movelytics build a movement feature that is specific to our brand?
Yes. The lab builds custom movement logic, custom keypoints and fine-tuned pose models. A project starts with a scoping call, continues with a proof of concept you can run on the target phones, then goes to production.
Do our users' videos go to Movelytics?
No. Inference runs on the user's phone. Movelytics collects usage metering only: the number of API calls, used for billing.
Does it work on older Android phones?
That is what we optimise for. In our preliminary benchmark of 17 November 2026, an internal build of the PoseTracker Model held 28.7 frames per second with the camera on a Huawei FIG-LX1 2017 (Android 9) over 10 minutes (slowest 10-second window: 22.1).
Fitness and wellness