Skin analysis
Measure redness, shine and texture from your camera — nothing leaves your device
- Free, no sign-up
- Runs in your browser
Camera
Line your face up with the guide, then take a reading. The camera starts only when you press the button, and stops the moment you leave this page.
Line your face up with the guide, then take a reading. The camera starts only when you press the button, and stops the moment you leave this page.
Nothing leaves your device. The frame is drawn to a canvas in your browser, measured there, and discarded.
No readings yet. Take your first one and it becomes the baseline everything after it is compared against.
What people use this for
- You cannot tell if anything is working. Skin changes slowly enough that memory is unreliable. A measurement taken the same way each time is not.
- Redness is the thing bothering you. Redness is hard to eyeball and easy to measure, because it separates cleanly in the right colour space.
- You want data without an app. No download, no account, and no image ever leaves your device.
What is actually being measured
Three numbers, each computed from the pixels inside a region you position yourself. Redness comes from converting the image into CIELAB colour space and reading the a* channel, which encodes the green-to-red axis independently of how bright the image is. That independence is the point: it is why a* is a far more stable redness measure than anything based on the red channel of RGB, which rises and falls with lighting.
Shine is estimated from the density of specular highlights — pixels that are both very bright and nearly colourless, which is what a reflection off sebum looks like. Texture is the local variance of brightness across the region: smooth skin varies little between neighbouring pixels, and rough or bumpy skin varies more.
These are image measurements, not clinical ones. They describe what the camera saw in that region under that lighting. They are meaningful compared against your own earlier readings and meaningless compared against anyone else's.
Everything happens on your device
The camera stream is drawn to a canvas in your browser and the arithmetic runs there. No frame is uploaded, no image is stored on a server, and there is no account. When you close the tab, the video is gone.
That is a deliberate design choice rather than a limitation we are making the best of. Face photographs are among the most sensitive data a person can hand over, and most tools of this kind send them somewhere. This one cannot, because there is no somewhere to send them to.
Lighting is the whole ballgame
The single biggest source of error is inconsistent light. Window light in the morning and a warm bulb at night will produce different numbers from identical skin, and the difference can easily exceed any real change. The tool checks overall brightness and warns you when a reading looks likely to be incomparable, but it cannot correct for it.
Get useful trends by standardising: the same place, the same time of day, the same light source, no makeup, and the same distance from the camera. Take a reading weekly rather than daily — daily variation is mostly noise, and watching it will tell you nothing except that skin fluctuates.
This is not a diagnostic device and it does not detect conditions. It quantifies three properties of an image. If something about your skin concerns you, that is a question for a doctor or dermatologist, and no amount of pixel arithmetic substitutes for someone looking at it.
Frequently asked questions
Is my photo uploaded?
No. The camera frame is drawn to a canvas in your browser, measured there, and discarded — the canvas is cleared and released as soon as the arithmetic is done. There is no server involved and no account. Only the resulting numbers are saved, and only in this browser.
Where are my readings kept?
In this browser's local storage on this device, as a short list of numbers with a date. They never leave the device and they will not follow you to another one. Clearing site data deletes them, so export the CSV if the series matters to you.
Why does it not detect my face automatically?
Landmark detection would mean downloading a machine-learning model of several megabytes from a third party. Positioning your face against a guide gets comparable regions without that cost, and the trade is stated rather than hidden. The presets and sliders are there so the boxes can be made to fit your face and your distance from the camera.
My numbers jump around between readings.
Almost always the light. Window light and a warm bulb produce different values from identical skin, and that difference can easily exceed any real change. The lighting check runs before you press the button for exactly this reason. Standardise the place, time and light source, and read weekly rather than daily.
Can this detect a skin condition?
No. It quantifies three properties of an image and nothing more. It is not a diagnostic device. If something about your skin concerns you, that is a question for a doctor or dermatologist.