Body-aware AI, private on your device

Hips Don't Die.
Your device just learned the moves.

Turn your iPad, laptop, or desktop into a private, body-aware movement coach. A gentle daily hip plan, live on-device pose tracking, yoga hold timers, and research-informed routines — without uploading your camera video.

Start today's gentle plan Open the camera coach

33-point pose AI
Experimental dual-model verification
Research-linked routines
No video upload
No account required

Tracks movement. Does not diagnose pain or prescribe treatment.

● Processed on this device

Neutral0 reps

A synthetic preview — no real person, no camera. The coach only uses your camera locally, after you tap start.

From camera to cue

Your webcam has rhythm. The AI handles the counting.

01
Pick your flow.
02
Set your starting range.
03
Let the local pose AI follow the movement.
04
Get repetitions, holds, and a private summary.

No universal “perfect pose.” The app follows the session range you select.

The AI stack

More than dots on a webcam.

Hips Don't Die combines a 33-landmark primary pose model, optional independent pose verification, temporal motion matching, uncertainty-aware counting, and adaptive browser acceleration. The entire movement pipeline runs locally.

01CameraLocal frames only
02MediaPipe Heavy/FullPrimary pose AI
0333 landmarksBody coordinates
04MoveNet verifierExperimental
05Uncertainty fusionConfidence-aware
06Temporal engineSession template
07Rep / hold stateDeterministic count
08Local feedbackOn your device

Primary pose AI

Adaptive Heavy/Full selection

Runs where it can

GPU, WebGPU, WebGL, and WASM fallbacks

Calibrated to you

Session-local motion templates

Experimental verifier

Dual-model disagreement detection

Relative geometry

3D landmark visualization

Local AI recap

Optional local generative recap

Open the AI Lab

Mobility + yoga

Yoga, but make it computer vision.

Five-Minute Hip Flow

A configurable standing sequence with marching, side movement, weight shifts, extensions, and a shallow hinge.

Supported Standing Yoga

Mountain, supported Tree, Warrior II, supported Chair, and an optional supported Crescent Lunge.

Chair, wall, or countertop? All excellent dance partners.

The research

Yes, the fun has receipts.

Exercise therapy has shown benefits for pain and function in research involving people with diagnosed hip osteoarthritis. Yoga research in older adults has reported improvements across measures such as balance, flexibility, strength, and physical function. Biomechanics research shows that different standing yoga poses create meaningfully different lower-body demands.

Hips Don't Die uses those findings to inform routine structure, supported options, and progression. They do not prove that this application — or any single movement — is right for every person or every hip.

The registry spans Cochrane reviews, a JAMA randomized trial, and analyses in The Lancet, BMJ, and the British Journal of Sports Medicine — 12 verified, DOI-linked studies in all, plus NHS and AAOS (orthopaedic) guidance at the movement level. Each one lists what it supports and, just as plainly, what it does not.

Cochrane systematic review and meta-analysis of randomized controlled trials

Fransen et al., 2014, Cochrane Database Syst Rev

Population: Adults with hip osteoarthritis
Result: Pooling trials of land-based therapeutic exercise found small reductions in hip pain and small improvements in physical function versus no exercise, with benefits sustained for a few months after treatment ended.
What it informs: Structuring the app around regular, gentle land-based hip movement
Framing consistent movement as generally helpful for people with diagnosed hip osteoarthritis
What it does not prove: That this app reproduces a studied exercise program
Large or curative effects — the measured benefits were small
Generalizing to acute injury or post-surgical hips

Fransen M, McConnell S, Hernandez-Molina G, Reichenbach S. Exercise for osteoarthritis of the hip. Cochrane Database of Systematic Reviews. 2014;(4):CD007912.

DOI: 10.1002/14651858.CD007912.pub2 · Last reviewed 2026-07-14

Systematic review and meta-analysis of randomized controlled trials

Sivaramakrishnan et al., 2019, Int J Behav Nutr Phys Act

Population: Older adults (mean age 60+), not recruited for any specific disease
Result: Compared with inactive controls, yoga was associated with improvements across physical-function measures such as balance, lower-body strength, and flexibility in older adults.
What it informs: Offering supported standing yoga as a general movement option for older adults
Emphasizing balance and gentle strength rather than intensity
What it does not prove: That yoga treats a specific hip diagnosis
That results in a general older-adult population apply to painful or post-surgical hips
That this app has been tested as an intervention

Sivaramakrishnan D, Fitzsimons C, Kelly P, Ludwig K, Mutrie N, Saunders DH, Baker G. The effects of yoga compared to active and inactive controls on physical function and health related quality of life in older adults — systematic review and meta-analysis of randomised controlled trials. International Journal of Behavioral Nutrition and Physical Activity. 2019;16(1):33.

DOI: 10.1186/s12966-019-0789-2 · Last reviewed 2026-07-10

Randomized, placebo (sham)-controlled clinical trial

Bennell et al., 2014, JAMA

Population: Adults with painful hip osteoarthritis
Result: A 12-week multimodal physical therapy program did not produce greater improvement in pain or function than a sham treatment, and the active group reported more mild adverse effects.
What it informs: Being honest that structured therapy is not guaranteed to help and can cause soreness
Keeping demand gentle and letting the user stop when something aggravates symptoms
What it does not prove: That physical-therapy-style exercise reliably relieves hip osteoarthritis
Any claim that this app improves pain or function
Ignoring that movement can cause discomfort for some people

Bennell KL, Egerton T, Martin J, Abbott JH, Metcalf B, McManus F, Sims K, Pua YH, Wrigley TV, Forbes A, Hinman RS. Effect of physical therapy on pain and function in patients with hip osteoarthritis: a randomized clinical trial. JAMA. 2014;311(19):1987-1997.

DOI: 10.1001/jama.2014.4591 · Last reviewed 2026-07-14

Read all 12 studies and their limits

Private by architecture

The camera sees frames. We don't.

Camera frames are processed in browser memory, converted into transient body landmarks, and discarded. The standard coach does not upload workout video or raw landmark streams.

Video stays local.

No identity tracking.

Local history is deletable.

No cloud AI is required.

MacBook camera
browser memory
on-device pose models
transient landmarks
frame discarded

AI with boundaries

No robot doctor energy.

The app tracks movement phases, repetitions, and calibrated holds. It does not identify the cause of pain, detect an injury, or determine whether a routine is medically appropriate.

Pricing

Start free. Keep the hips moving.

Keep Moving

Free

Five-Minute Hip Flow
Supported Standing Yoga
Live pose tracking
Repetitions and hold timers
Dual-model verification when supported
Flow Beat
Three local summaries
No account required

Move those hips

Hips Forever

$39 once

Founding access

Everything in Keep Moving
Custom local routines
Unlimited local summaries
JSON and CSV exports
Configurable order, repetitions, holds, and rests
Custom Flow Beat controls
Local AI recap controls
Expanded 3D Motion Lab
Future routine-template packs

See founding access

Questions your hips may have.

Is this really AI?

Yes. The coach runs pretrained on-device machine-learning pose models (MediaPipe Pose Landmarker, and optionally MoveNet Thunder as a verifier). When your browser has a local foundation model, an optional recap can also be generated on-device. The deterministic counting logic on top is transparent by design.

What runs on my device?

The pose models, the movement pipeline, the visualizations, and any optional local recap all run in your browser. Model files are served from this site's own origin and cached locally after first use.

Does my video leave my computer?

No. Camera frames are processed in browser memory and discarded. The standard coach does not upload workout video or raw landmark streams anywhere.

What is dual-model verification?

An experimental second pose model (MoveNet Thunder) independently checks selected shared landmarks. When the two models disagree substantially, the app pauses counting rather than guessing. It is not clinical validation.

Does the app diagnose hip pain?

No. It tracks movement phases, repetitions, and calibrated holds. It does not identify the cause of pain, detect injury, or decide whether a routine is medically appropriate for you.

Are the routines backed by research?

Routine structure is informed by peer-reviewed research on exercise, yoga, and hip biomechanics. The Research page lists each study, what it supports, and what it does not prove.

Do I need to be flexible?

No. Every routine calibrates to the range you show it during practice, and it never asks you to go deeper.

Can I use a chair or wall?

Absolutely. Chair, wall, or countertop? All excellent dance partners.

Does it work in Safari?

The core coach targets modern Chromium and WebKit browsers. Some acceleration and experimental features vary by browser; the app selects the best available path and tells you which mode is active.

Do I need an account?

No. The full local coach works with no account and no environment configuration.

What is the local AI recap?

An optional, experimental post-session summary generated by your browser's local language model when available. It only ever sees non-medical session totals, and there is always a deterministic fallback.

Your hips called

They want a software update.

Start with today's gentle plan — one tap, no account — or open the camera coach and let the on-device AI handle the counting.