Form correction has traditionally required a personal trainer standing beside you, watching every rep. ENLIVEN's ML Vision system brings that level of analysis to your phone camera.
The Technology
Our pose estimation model uses a lightweight neural network optimized for mobile devices. It tracks 33 anatomical keypoints in real time at 30 frames per second:
- Head, shoulders, elbows, wrists
- Hips, knees, ankles
- Spine alignment points
- Foot placement markers
How Form Analysis Works
For each exercise in our library, we've defined the biomechanically correct movement pattern. During your set, ML Vision compares your actual joint angles and positions against this reference pattern.
When a deviation exceeds the safety threshold, you receive immediate audio feedback: "Keep your back straight" or "Drive through your heels." This happens in real time — not after the set is over.
The Ghost Overlay
After each set, ENLIVEN generates a "ghost overlay" — a visual comparison of your form against the ideal movement path. Think of it like the racing line in Formula 1: you can see exactly where your movement diverged from optimal.
This isn't about perfection. It's about progressive improvement. The system tracks your form scores over time, showing you which exercises are improving and which need attention.
Rep Counting
ML Vision also handles automatic rep counting with 95%+ accuracy. No more losing count during a difficult set. The system recognizes the full range of motion for each exercise and counts completed reps, partial reps, and failed attempts separately.
Privacy First
All real-time processing happens on-device. Your camera feed is never uploaded to our servers. Post-set analysis uses anonymized skeleton data only — no video is stored unless you explicitly choose to save it for your own review.