What this tool does
The model looks at a standing photo the way a cautious coach would: waist relative to shoulders, muscle separation, and how clothing hides shape. It returns a range and a confidence, not a lab percentage. It does not replace a tape, a DEXA, or a clinician.
How it works
You upload a JPEG, PNG, or WebP. The image is sent to a vision model on the server, scored as structured JSON, and discarded. Sex on the form is visual context. If the photo is a child, a group, or too covered to judge, the result says so instead of inventing a precise number.
How to interpret the result
Treat the midpoint as a conversation starter. If last month’s photo under the same light and clothes was clearly leaner, that trend matters more than a 1.2-point change. Confirm with the Navy tape method when you can repeat landmarks.
Example
A front-relaxed photo in even indoor light, shorts and a fitted shirt, might print something like 16–22% with medium confidence. The same body in a baggy hoodie or a bathroom selfie with overhead shadows should drop confidence and widen the range.
Common mistakes
- Flexing, sucking in, or using a filter and calling it data.
- Comparing a morning tape to an evening party photo.
- Using a single estimate to crash calories.
- Uploading someone else’s photo.
When this estimate may be off
- Baggy clothes, heavy shadows, or a seated/cropped shot.
- Very high muscle mass or atypical fat distribution.
- You need a clinical composition test.
- The file is a screenshot of another app’s result.