How accurate is a body fat estimate from one photo?

A single-photo estimate typically carries ±4–5% error compared to DEXA. It's good enough for a directional read on lean / average / overweight, but not for tracking small week-over-week changes. The GainFrame iOS app uses multiple angles per check-in to cut that error roughly in half.

Is my photo stored or used to train AI?

No. Your photo is sent to Google's AI service for the analysis call only. It's not stored on GainFrame's servers and not used to train AI models. The image is processed and discarded.

Why am I limited to one estimate per day?

The web tool is free and rate-limited to one estimate per day per user to keep AI compute costs sustainable. The GainFrame iOS app removes the limit and adds tracking, scoring, and comparison features.

What kind of photo works best?

A clear, well-lit photo where your torso, arms, and at least the upper part of your legs are visible. Tight or revealing clothing helps. Baggy clothing, side angles, and face-only shots produce poor estimates.

Does this work for women?

Yes. The visual reference ranges shift between male and female bodies, so selecting your sex before submission improves accuracy.

Free AI Body Fat Estimator from a Photo

Your photoStep 1 of 2

Drop a photo or click to upload

JPEG · PNG · HEIC — up to 10 MB

ReferenceImproves accuracy

Privacy: photo sent to Google's AI for the analysis call only · Not stored, not used to train models.

How the scan works.

Step 01

Upload one clear photo

Front-facing, torso and limbs visible. Tight clothing or shirtless gives the AI more visual information to read.

Step 02

Pick a reference

Body fat ranges differ meaningfully between male and female physiques. Selecting yours improves the estimate.

Step 03

Get a number plus a confidence read

A single percentage with confidence band. Single-photo carries ±4–5% error — directional, not week-to-week.

How accurate is body fat from a photo?

A 2025 peer-reviewed study in npj Digital Medicine tested AI-2D photo body fat estimation against DEXA across 1,273 adults. The AI method produced a Concordance Correlation Coefficient of 0.98 with DEXA — higher agreement than bioelectrical impedance smart scales (which typically score 0.91–0.92) and within the range of clinical imaging methods.

That said, photo-based estimation has real limits. A single photo carries ±4–5% absolute error vs. DEXA; lighting, posture, and clothing all affect the read. The technique is best used for trend tracking (same setup, same time of day, same wardrobe) rather than as a one-shot diagnostic. For tighter accuracy, the U.S. Navy tape-measure method uses neck/waist/hip circumference and lands around ±3% — more effort, no AI involved. For the gold standard, see our DEXA alternatives breakdown.

The GainFrame iOS app uses a multi-angle Precision BF model that combines front and side photos, cutting single-photo error roughly in half. The web tool you just used is the single-photo version — fast, free, and a good directional read.

More ways to measure.

U.S. Navy tape-measure calculatorSlightly more effort — neck, waist, hip — but tighter ±3% accuracy and zero AI involved.

Open →

Body Fat VisualizerDrag two sliders to see photorealistic references for every body fat percentage (men 8–33%, women 18–42%).

Open →

FFMI CalculatorOnce you have your body fat %, plug it in here for a fat-free mass index — the natty-limit metric.

Open →