Camera
Setup Guide
How to Get Accurate Results from the Face Shape Detector
The difference between a useful face shape result and a confusing one usually comes down to one thing: the photo. The AI is only as good as the input it receives. A slightly tilted head, a shadow across the jaw, or hair covering the hairline can shift the landmark detection enough to produce the wrong shape classification entirely.
This guide covers how to take or capture a photo that gives the AI what it needs — step by step, with the specific details that actually matter and the mistakes that most commonly cause inaccurate results.
What the AI Is Actually Measuring
Understanding what the face shape detector is doing helps explain why photo quality matters so much. The AI uses facial landmark detection — it maps specific anchor points on your face, including the corners of your jaw, the outer edges of your cheekbones, your temples, and the center of your hairline. From these points, it calculates four measurements: forehead width, cheekbone width, jawline width, and face length.
The ratios between these measurements determine your face shape classification. This means any factor that obscures, distorts, or shadows one of these anchor points affects the final result. A shadow under the jaw changes the chin measurement. Hair over the temples changes the forehead width reading. A tilted head changes every measurement simultaneously.
The Four Measurements the AI Uses
- →Forehead width — measured at the widest point, halfway between hairline and brows
- →Cheekbone width — measured from outer eye corner to outer eye corner across the cheekbones
- →Jawline width — measured from the chin tip to the jaw angle, doubled
- →Face length — measured from center hairline to chin tip
Using the Camera Feature: Step by Step
The camera feature lets you capture a photo directly in your browser without searching through existing photos. The live view also lets you confirm your positioning and lighting before capturing — which is its main advantage over uploading.
- 1
Open the detector and select Camera
Navigate to the Face Shape Detector and click the "Use Camera" button. Your browser will request camera permission — click Allow to proceed. If you accidentally deny permission, see the troubleshooting section below for how to re-enable it.
- 2
Set up your environment before positioning
Before framing your face, check the lighting in the live view. You want even light on your face with no strong shadows — particularly no shadows under the jaw or across one side. Move to face a window or a bright light source in front of you, not behind you.
- 3
Position your face in the guide
Center your face within the on-screen oval guide. Your entire face — from hairline to chin — should be visible. Pull all hair back from the face and forehead so the hairline and temples are fully exposed. Keep the camera at exact eye level: not above (looking down at you) and not below (looking up at you).
- 4
Level your chin and hold still
Keep your chin level — not lifted, not tucked. A lifted chin flattens the jaw measurement and makes your face appear shorter; a tucked chin extends it. Look straight into the lens with a neutral, relaxed expression. No smiling — a smile shifts the jaw and cheek landmarks.
- 5
Capture and review
Click the capture button. If the result doesn't look right or the lighting was off, use the retake option before submitting for analysis. The live view lets you take multiple attempts — use it. Submit only when the photo is well-lit, centered, and hair-free.
Lighting: The Single Biggest Variable
Lighting has more impact on result accuracy than any other single factor. The reason is specific: the AI detects facial landmarks by reading contrast and edge definition. Poor lighting eliminates the contrast the model needs to locate the jaw corners, cheekbone edges, and hairline boundary accurately.
What Good Lighting Looks Like
Even, diffuse light that illuminates the entire face without creating shadows. The ideal setup is to face a window with natural daylight coming in — not sunlight streaming directly in (which creates harsh shadows), but bright daylight from an overcast sky or a north-facing window. Indoors, a ring light or a lamp positioned in front of you (not to the side) achieves a similar result. The goal is a face with no shadowed zones — you should be able to see both sides of your jaw clearly in the preview.
Good Lighting Sources
- ✓Overcast daylight facing a window — diffuse, even, no harsh shadows — ideal for landmark detection
- ✓Ring light positioned in front — consistent, even illumination that matches what the model was trained on
- ✓Bright indoor overhead + front lamp — combining overhead and frontal light fills in under-jaw shadows
- ✓Softbox or diffused studio light — professional-grade evenness; if available, this is the best option
Lighting to Avoid
- ✕Backlighting (window behind you) — silhouettes the face and destroys all landmark contrast; the single most common mistake
- ✕Overhead lighting only — creates deep shadows under the jaw and nose that distort the chin measurement
- ✕Side lighting (one side brighter) — puts half the face in shadow, making the cheekbone and jaw measurements asymmetric
- ✕Very dim or night conditions — noise and low contrast affect landmark accuracy in all directions
"The AI detects landmarks by reading contrast and edge definition. Poor lighting doesn't just make the photo look bad — it removes the signal the model depends on."
Positioning Details That Change Results
Each positioning variable below affects a specific measurement. Understanding which measurement is affected helps explain why the guidance matters beyond general advice to "hold the camera straight."
Camera height
The camera must be at eye level — not above or below. A camera angled down at you compresses the forehead and exaggerates the chin length, pushing the result toward oblong. A camera angled up at you does the opposite — it shortens the chin and widens the jaw, pushing toward a round or square classification. Even a modest angle of 10–15 degrees is enough to shift the measurement by a noticeable margin.
Chin angle
Keep the chin level and parallel to the floor. A lifted chin makes the face appear shorter and flattens the jaw measurement. A tucked chin creates the appearance of a longer face and changes the jaw angle the AI reads. If you're unsure whether your chin is level, look at your ears in the preview: they should be at the same horizontal position as each other.
Head rotation
Face directly forward — no turning left or right. Even a small rotation exposes one side more than the other, making the face appear asymmetrically wider on the side facing the camera. This directly affects cheekbone and jaw width readings.
Head tilt
Keep the head vertical with no tilt left or right. A tilt makes one side of the jaw appear lower than the other, complicating the landmark model's ability to find the true jaw corners accurately.
Distance from camera
Fill the frame with your face — not so close that the edges of your face are cut off, and not so far that your face is a small portion of the image. The face should occupy roughly 60–70% of the frame height. Too small and landmark precision drops; too close and the edges of the forehead or jaw may fall outside the detection zone.
Camera Feature vs. Uploading a Photo
Both methods produce accurate results when the input photo is good. The choice between them comes down to what each method makes easier.
Camera vs. Upload — Comparison
| Camera Feature | Upload a Photo | |
|---|---|---|
| See lighting before capture | ✓ Yes — live preview | ✗ No — photo already taken |
| Verify positioning | ✓ Yes — real-time guide | ✗ No — fixed image |
| Retake instantly | ✓ Yes — as many times as needed | △ Requires a new photo |
| Use existing photos | ✗ No | ✓ Yes |
| Works without a camera | ✗ No | ✓ Yes |
| Best for first-time use | ✓ Yes | △ Depends on photo quality |
For first-time users, the camera feature is generally the better starting point because the live preview makes it easy to confirm lighting and positioning before committing to a capture. For users who have a recent, well-lit, front-facing photo that was taken under good conditions, uploading that photo is equally valid and slightly faster.
Common Issues and How to Fix Them
Camera permission denied
If you accidentally clicked "Block" when prompted, the browser won't ask again automatically. In Chrome: go to Settings → Privacy and Security → Site Settings → Camera, find faceshapedetector.app in the list, and change the permission to Allow. In Safari: go to Safari → Settings for This Website and enable camera. Then refresh the page.
Camera not detected or showing a black screen
First check whether another app is currently using the camera (video calls, photo booth apps). Close those apps and try again. If the problem persists, restart your browser. On some devices, the browser may need a full restart to re-detect the camera after it's been released by another application.
Result seems wrong or unexpected
The most common cause is a photo that violates one of the positioning or lighting principles above — particularly camera angle (not at eye level) or a shadow across the jaw. Retake under better conditions. If the result consistently shows a shape that doesn't seem right, compare your measurements manually against the shape profiles in our full face shape guide.
Image is blurry or low quality
Clean the camera lens with a soft cloth — this is more effective than it sounds. Make sure you're holding still at the moment of capture. On older devices, moving to a brighter environment often improves sharpness because the camera can use a faster shutter speed with more available light.
Face not detected at all
This typically means the face is too small in the frame, too dark, partially obscured, or at a significant angle. Move closer to the camera until your face fills about 60–70% of the frame, check lighting, remove any accessories (hats, large earrings), pull back hair, and ensure you're facing directly forward.
The Complete Checklist Before You Capture
✓ Ready to Capture
- ✓Lighting is even on both sides of face — no shadows across the jaw or on one cheek
- ✓Camera is at eye level — not angled up or down
- ✓Chin is level and untilted — ears at the same height in the frame
- ✓All hair is pulled back — hairline, temples, and full jaw are visible
- ✓Facing directly forward — no rotation left or right
- ✓Neutral expression — no smile, no squinting
- ✓Face fills ~60–70% of frame height — close enough for landmark precision
✕ Not Ready Yet
- ✕Backlight or window behind you — silhouettes the face completely
- ✕Camera above or below eye level — distorts face length and jaw measurement
- ✕Hair covering hairline or temples — changes forehead width reading
- ✕Head tilted or rotated — affects every measurement simultaneously
- ✕Hat, headband, or large earrings — can obscure landmark points
- ✕Smiling or making an expression — shifts jaw and cheek landmark positions
- ✕Face too small in frame — reduces landmark precision