Your physique,
precisely
measured.
LeanBody reads your body fat percentage, lean mass, and FFMI from four photographs (front and back, flexed and relaxed), scored against the NHANES 2017–20 DXA reference cohort.
You’re leaner than
the mirror says.
Women carry more essential fat by design, so a lean, athletic woman often reads around 22%, not 12%. If a body fat number has ever scared you, it was probably normal and healthy the whole time. LeanBody gives you one honest read, kept private, with nothing stored on an account.
No black box.
Here’s the method.
- i.Four photographsFront and back, flexed and relaxed, from your phone. They go straight to the model.
- ii.A frontier vision model reads themThe same class of multimodal model behind tools you already use, given a fixed protocol and all four photos at once.
- iii.Scored against NHANES DXA dataYou get a percentile against the CDC reference cohort, plus lean mass and FFMI. Then the photos are deleted.
Free tools vibe a number.
For a one-off rough check, a free tool is fine. This is for tracking, where the method has to stay fixed or the number is just noise.
- Consistency
- A free one-off, or asking ChatGPT, hands you a fresh guess each time with nothing to anchor to. LeanBody uses the same four-photo protocol and benchmark every scan, so the number holds meaning across months.
- A real benchmark
- Your result is a percentile against NHANES DXA data, not an unanchored number.
- Privacy by deletion
- No account, no database. Photos are analysed, then deleted. They never sit in a chat history.
- A series, not a subscription
- One reading is $5. A series of six is $20, the same fixed method each time, so you can watch the number move month over month. No recurring charge, no upsell, refund on request.
Photos vs. DXA, in the literature.
Estimating body fat from a camera isn’t a gimmick. Peer-reviewed studies, from single smartphone photos to 3D optical scans, have validated the approach against DXA, the gold standard, and found it beats the bioimpedance scales most people already distrust.
Smartphone-photo body composition matched DXA with ~2% mean error and near-zero bias, and beat every bioimpedance and BodPod method tested.
In the Shape Up! Adults study, 3D optical body scans tracked DXA fat mass at R² 0.94 (~2.9 kg error), with test-retest precision close to DXA's own.
The honest caveat. Other work finds photo methods reliable but biased depending on the model and conditions, and none of it replaces a DEXA. These studies validate the approach with purpose-built models, not LeanBody’s own numbers. We apply the same idea, benchmarked to NHANES DXA, and we don’t pretend to be a scan.