How HexaFlow built SwingSmith: upload one swing from a phone and get a 10-position breakdown, 50+ metrics, and the single fix worth practising next.
Shipped on iOS and Android, rated 4.3 on the App Store, turning ordinary phone video into one measurable practice priority.
See it liveswingsmithpro.comThe club comes steeply from outside the target line, forcing early extension just to make solid contact.
Generic advice can sound smart and still leave a player guessing which of a hundred possible faults is actually theirs — and a phone video on its own tells them nothing they can act on.
Real coaching doesn't scale. An experienced coach names the fault in seconds, but that expertise is expensive, limited by geography, and impossible to bring to every practice session.
It is easy to return fifty numbers. The value is in returning one instruction a golfer can take to the range.
The harder product problem was restraint
The golfer records a face-on or down-the-line swing on an ordinary phone and uploads it in a few taps.
No launch monitor, no studio, no specialist hardware — the constraint that has kept swing analysis locked inside coaching facilities. Everything that follows has to work from that one video.
The engine breaks the motion into ten checkpoint positions, reads 50+ metrics across path, tempo, head stability, spine inclination, shoulder tilt, hip turn and transition, and scores the swing. Drag through it.
Try itScrub the bar or pick a frame. Measurements come from this page's illustrative swing model and the frame scores are sample values — not the product's engine.
Numbers alone don't change a movement. The analysis is drawn back onto the golfer's video, and set beside pro checkpoints at matching positions — so the gap is something they can see.
| At P6 | You | Pro |
|---|---|---|
| Hand height | 0.59 | 0.54 |
| Shoulder turn | 32° | 32° |
| Hip turn | -4° | -14° |
| Head position | 0.01 | -0.01 |
Try it Step through positions and switch overlays. Both golfers are drawn from the illustrative model.
Rather than reporting everything, the system names the one change most likely to improve the next range session.
And it explains the causal chain — what the fault is, what it forces the body to do — so the golfer understands why it matters rather than just being told to fix it.
Every reading here is accurate. Taken together they still leave a golfer guessing which of them is actually theirs to fix — and which to ignore.
Try itSwitch between the two. Readings are sample values; the fault and its practice focus are the ones the product's swing sequence uses.
Each priority fix maps to a specific drill. Ask Brody, the in-app AI coach, answers follow-up questions grounded in the golfer's current analysis and their previous uploads — so the answers are about their swing, not golf in general.
Progress is tracked by comparing matched checkpoints across uploads, so a golfer can see whether a fault is genuinely improving.
Alright, let's get straight to it.
Main diagnosis: a steep, over-the-top path. The club comes from outside the target line, and you're extending early just to make contact.
Your one thing: from the top, feel the club drop into the slot behind you, like pulling a bell rope, then let it approach from the inside.
Do not focus on: the early extension itself. It's the symptom; the path is the cause.
Priority fix: steep / over-the-top path
Bell rope: drop the club into the slot, approach from the inside
Sample conversation and history.
Try it Ask a question and watch which part of the swing memory the answer uses.
Everything above is a recreation. These are captures from the app on iOS.
Swings upload from the native apps to cloud storage, queue for inference, pass through pose estimation and the P10 checkpoint classifier, and return a structured analysis that drives the overlays, the priority fix, and the context Ask Brody reasons over.
Every analysis is kept, so matched checkpoints can be compared across uploads to show whether a fault is really improving.
Native iOS and Android apps with a free starter analysis and tiered subscription plans, backed by the computer-vision pipeline and the coach above.
ServicesComputer Vision AI & Machine Learning Mobile Application Development NextGen Technologies
Live on iOS and Android with a 4.3 App Store rating
Swing analysis from ordinary phone video, with no specialist hardware
One priority fix per swing instead of an undifferentiated metric dump
Follow-up coaching grounded in the golfer's own upload history
Progress measurable across sessions through matched checkpoints
A walkthrough video in, a removals quote in seconds
The customer films their own home. Computer vision identifies and counts every object, estimates dismantling, handling and transport time, and returns a priced quote.
Three role modules — school, teacher and student — with computer-vision proctoring
A multi-tenant platform with three distinct role modules — school administrator, teacher and student — backed by computer-vision proctoring, a randomised question engine and automatic certification on pass. We also produced the entire curriculum that runs inside it.
Insurance document capture and submission
A mobile app that scans, cleans up and submits insurance documents in three taps.