Headshot Quality Checker

What the score means in plain English

The overall number is a weighted average — sharpness and lighting count 25% each, framing 20%, resolution and background 15% each. The weights reflect what hiring managers and recruiters cite most often when they reject a profile photo: out-of-focus first, dim/dark second, awkwardly framed third. Resolution and background are the easy fixes, so they count less in the weighted score.

Why each metric matters

• Resolution. LinkedIn displays 400 px on desktop. Anything below that gets upscaled by the platform and looks soft. Retina displays double the requirement — 800 px is the safe upload minimum, 1200 px is preferred for any print use.
• Sharpness. The single most-cited "why this photo is bad" reason. Almost always either focus (camera focused on your shoulder, not your eye) or motion (handheld at slow shutter). Tap-to-focus on the eyes is the universal fix.
• Lighting. Even good cameras can't recover from a bad lighting setup. Indirect window light beats every other option. Direct flash is the worst.
• Framing. The face should fill 35-60% of the frame height, with about 10% margin above the head. If your face is too small, the platform's thumbnail crop makes you look like a speck. Too tight and you look claustrophobic.
• Background. Solid wins. Even a slightly busy background pulls the viewer's eye off your face — which is the entire point of a headshot.

How the heuristics work under the hood

The Laplacian variance metric is the standard cheap focus measure used in microscopy and photography QA — it computes a Laplacian (second-derivative) filter on a grayscale downscale, then takes the variance. Sharp images have lots of high-frequency edges, which means high variance. Blurry images have flat regions and low variance. Threshold values are calibrated against a corpus of 500+ LinkedIn profile photos.

The face detector uses the browser's experimental window.FaceDetectorAPI where available, falling back to a conservative neutral score where it isn't. The background metric samples four corner regions of the image and measures color variance between them — a uniform background gives low variance.

If your score is bad

Don't try to fix every metric individually. A photo scoring 45 on sharpness and 50 on lighting isn't going to become a great photo with a Photoshop pass — the right call is to re-shoot, or generate a fresh one in the AI Studio. For photos scoring 70+ overall with one weak metric, the targeted fix in the result card is usually enough.

What this tool doesn't cover

We don't score expression, outfit, hairstyle, or "professional vibe." Those are subjective and culture-dependent. We also don't detect AI-generated photos — we score technical quality regardless of how the photo was made. If you want a second opinion on whether a photo looks like an AI render, read our piece on whether recruiters can tell AI headshots.