Personal Color Card
Upload a photo for a full personal color analysis — your season, palette and a side-by-side draping test — then save the whole result as one shareable card. The face is found automatically and your skin is measured, not guessed from a quiz.
Drop a photo here, or click to choose one
That is the only step. Your face, eyes and hair are found for you, and the result appears straight away.
It is opened and measured inside this browser tab, on your own device. Nothing about it reaches a server, and it is gone the moment you close the page.
One thing matters more than everything else
Get something white or grey in the shot — a sheet of paper, a plain wall. Across ordinary indoor and outdoor lighting the same face measures anywhere from ITA −7° to +88°, which is every skin-depth group there is. A known neutral in frame collapses that to under a degree, because it pins down both the colour of the light and the exposure.
Then: indirect daylight, no makeup, no filter or beauty mode, hair back off the face.
What personal colour actually is
It is one narrow, practical idea: some colours make your skin look even and rested, and others drop grey or yellow into it and pull attention to shadows and lines. Hold two shirts under your chin and one of them makes you look like you slept well. That is the whole phenomenon.
It is not a personality type, it is not related to star signs, and it does not say anything about you beyond how light bounces off your face. It also does not forbid you from wearing anything. A palette is a shortlist for decisions you make in a hurry — in a shop, in front of a mirror at 8am — not a rulebook.
The reason it works at all is that your skin has its own colour, and fabric next to it either agrees with that colour or fights it. Everything below is about reading which is which.
Undertone — which pigment leads
Cool
44°
boundary
51.4°
Neutral
56.3°
boundary
61.2°
Warm
70°
Real skin covers a surprisingly narrow band. The difference between the two ends is smaller than people expect, which is exactly why guessing from memory does not work.
Depth — how light or deep
Very light
ITA 63°
Light
50°
Intermediate
34°
Tan
12°
Brown
−18°
Dark
−49°
A separate axis. Someone very light and someone tanned can both be warm — depth and undertone are independent.
The twelve types, and what each one wears
Four families, each split three ways. Spring and Autumn are the warm side, Summer and Winter the cool side; within a family the split is by how light, how deep, or how muted the colouring is.
Spring
Light Spring
True Spring
Bright Spring
Summer
Light Summer
True Summer
Soft Summer
Autumn
Soft Autumn
True Autumn
Deep Autumn
Winter
Deep Winter
True Winter
Bright Winter
Top row of each card is the palette, bottom row its neutrals. Hover any swatch for its name.
Reading your result
You get a season name, a match percentage, and a runner-up. The runner-up is not filler — it is the honest part.
- A clear gap (10 points or more) means your colouring sits comfortably in one type. Use that palette.
- A close call (under 6 points) is flagged as borderline, and it is genuinely common. You are between two types. Take the colours the two palettes have in common and treat those as your safest ground — they will be the ones that work regardless.
- The full ranking at the bottom shows all twelve. If your season seems wrong, look at which types scored close behind and try their palettes in the draping panel.
The four bars show why you landed where you did: how warm or cool your skin reads, how light or deep your colouring is overall, how vivid your hair and eyes are, and how much contrast there is between them. The dot is you, the notch is the ideal for your season. A dot far from its notch on one bar tells you exactly which way you differ from the textbook version of your type.
The draping test is the part to trust
Draping is what a professional consultation actually consists of: fabric held under the chin, one colour at a time, watching the face. It is also the one part of that process a screen can reproduce honestly, which is why it is on this page.
Use it properly and it will tell you more than the label does. Look at the skin, not at the colour. With a shade that suits you, shadows under the eyes recede, redness settles down, and the face looks more even. With a wrong one, the skin picks up a grey or yellow cast, under-eye circles deepen, and any lines around the mouth get more visible.
A warm complexion, two grounds
From the palette
From the avoid list
The circle is the identical colour in both. Against rust it looks warm and solid; against icy pink it turns sallow. Nothing changed but what is next to it.
A cool complexion, two grounds
From the palette
From the avoid list
Same effect, reversed. This is simultaneous contrast — a real property of vision, not a styling opinion.
Click through your palette and then click one from the avoid list. If you cannot see a difference between them, your colouring is probably fairly neutral and the whole question matters less for you than the internet suggests — which is a legitimate outcome, not a failure.
What to do with it
Only what touches your face matters
A collar, a scarf, a knit, a hat. Trousers and shoes are far enough away that the effect is negligible. If you take one thing from this page, make it that — it turns a palette from a wardrobe overhaul into a rule about tops.
Your neutral is the useful bit
The palette colours are fun; the neutrals are what you will actually buy. Knowing whether your version of "a plain top" is ivory, soft white, camel or true black is worth more than knowing your season name.
Metals, glasses, hair
The same logic covers frames and jewellery, which sit right against the face all day. Hair colour shifts your result more than anything else you control — dyeing it changes the contrast and depth axes, so a big change means re-running this.
Wear the colour you love anyway
If a shade is on your avoid list but you like it, move it away from your face — a bag, a skirt, a jacket worn open over something in your palette. Nothing here is worth giving up a colour you enjoy.
Things that are commonly said and are not true
- "Check whether your veins look blue or green." Veins look blue-green through skin because of how tissue scatters light — shorter wavelengths bounce back to your eye while longer ones are absorbed. It depends on how deep the vein sits and how translucent the skin above it is, not on undertone. Real optics, wrong question.
- "Everyone has one true season." Colouring is continuous. The twelve types are buckets placed on a smooth range, and plenty of people sit on a boundary. That is why this page shows the runner-up and the gap.
- "Gold or silver tells you your undertone." It is a reasonable draping test in itself, but it is the same test as everything else here, not an independent confirmation of it.
- "A tan changes your season." It changes your depth, not your undertone. Your palette may need to go a little richer in summer; the warm-or-cool answer does not move.
- "You can do this from any selfie." This is the one that actually matters, and it is why the tool nags you about lighting. Makeup, filters, beauty mode and coloured lighting all change the thing being measured.
Getting a photo that gives a real answer
The single biggest improvement you can make is to hold a sheet of white paper next to your face and tell the tool that is what it is. That one step is the difference between measuring you and measuring your light bulb — there is a section on the other tab with the numbers, and they are not subtle.
- Light: near a window, out of direct sun. Not under a warm bulb, not in a room with a strongly coloured wall.
- Face: clean, no makeup — foundation is literally designed to change your undertone.
- Camera: no filter, no beauty mode, no portrait smoothing. Check your phone is not applying one by default.
- Framing: front-on, hair back off the face, face filling a decent part of the frame.
Everything after that is automatic — the face, the three colours and the lighting correction are all found for you. If no face is found at all the tool says so, and a better-framed photo is the fix.
What is measured, and what is convention
Most pages answering "am I warm or cool?" do it with a questionnaire. Those questions are proxies for a physical quantity that can simply be measured, and measuring it is what this page does. The honest split is worth stating plainly, because almost nobody else does:
Measurement — reproducible
Your skin's L*, a*, b* in CIELAB; its ITA angle; its hue angle and chroma; the lightness gap between skin, hair and eyes. Anyone with the same photo and the same maths gets the same numbers.
Convention — a styling system
The step from those numbers to "True Autumn" comes from the Sci\ART and Color Me Beautiful tradition. It is a useful shorthand used by working stylists, but no peer-reviewed literature establishes it, and different practitioners would draw the boundaries differently.
That is why the tool shows the raw numbers, the four axes, and how every type scored — not just a verdict.
ITA°, the number dermatologists use
The Individual Typology Angle, introduced by Chardon and colleagues in 1991, is the standard instrumental measure of constitutive skin colour, used in dermatology and photoprotection research:
L* is lightness and b* is the yellow–blue axis, so ITA combines how light the skin is with how yellow it is. Two people can share an L* and land in different groups, which is the distinction a lightness reading alone throws away.
| ITA range | Group | Typical sun response |
|---|---|---|
| above 55° | Very light | Burns easily, rarely tans |
| 41° – 55° | Light | Burns, then tans slightly |
| 28° – 41° | Intermediate | Occasionally burns, tans steadily |
| 10° – 28° | Tan | Rarely burns, tans readily |
| −30° – 10° | Brown | Very rarely burns, tans deeply |
| below −30° | Dark | Deeply pigmented, does not burn |
Undertone is a hue angle
All human skin colour comes from the same short list of pigments, dominated by melanin — pushing toward yellow-brown — and haemoglobin, pushing toward red. Undertone is which of the two leads, and in CIELAB that is the hue angle, h° = arctan(b*/a*).
Because it is the same two pigments, skin occupies a narrow slice of the colour wheel. Sweeping the plausible facial region of CIELAB (L* 45–80, a* 6–18, b* 10–26) for colours that are in sRGB gamut and pass the skin detector gives 226,135 valid skin colours spanning 29° to 77°, median 56.3°. The cool/neutral/warm boundaries this page uses are the terciles of that distribution — 51.4° and 61.2°.
They are still a convention. Nature has no line where warm becomes cool, which is why your raw angle is always shown on the full scale beside the label.
The lighting problem, measured
A camera records light reflecting off you, which is the product of your skin and the illuminant. To find out how much that matters, one complexion was re-lit across the Planckian locus from 2700 K to 9000 K at exposures from −1.5 to +1.5 EV:
One complexion, six ordinary shooting conditions
Warm bulb
2700 K → Tan
Neutral LED
4000 K → Intermediate
Overcast
6500 K → Light
Deep shade
9000 K → Very light
Under-exposed
−1.5 EV → Brown
Over-exposed
+1.5 EV → Very light
Every square is the same skin — #e0b48c under overcast daylight — re-lit. The tool would classify these as six different people.
| Correction | ITA° spread | Groups the same face landed in |
|---|---|---|
| None | 95.0° | All five, from Brown to Very light |
| Neutralise the channels only | 104.5° | All five — no improvement whatsoever |
| Reference of known lightness | 0.6° | One, and the correct one |
The middle row is the important one. A plain white-balance step removes the colour cast and leaves exposure untouched — and ITA° is dominated by lightness, so it achieves nothing. Correcting to a reference whose lightness is known pins exposure as well, and collapses the spread to under a degree. The undertone reading follows: 50° of spread becomes 1.9°.
That is the entire argument for getting a sheet of paper in the shot. Without one, undertone and contrast still carry information because they are ratios, but the depth reading measures your lighting rather than you — and the result says so rather than quietly reporting a number.
How the face is found
Detection is Google's BlazeFace short-range model running in the MediaPipe vision WASM runtime, loaded only when a photo is opened. It replaced a hand-written skin-colour and local-contrast heuristic that worked on synthetic faces and fell apart on real photographs.
- Reference correction first, in linear light, so the detector and every measurement see the same corrected pixels.
- BlazeFace returns a box and six keypoints — both eyes, nose tip, mouth and both ear tragions. The largest face wins if several are in frame.
- Everything is scaled off the distance between the pupils, not off the face box. Warm blonde hair passes a skin detector and inflated the old box by a third, which put the cheek samples in the jaw shadow.
- The pupil is refined before the iris is read. BlazeFace marks the centre of the eye, not the pupil, and the few pixels between them are enough to pull sclera into the sample.
- Iris from an annulus, not a disc — a disc averages in the black pupil and the white of the eye and returns a grey belonging to nobody. The dark and light quarters are dropped, removing the lash shadow and the catchlight.
- Cheeks are median-filtered with the tails trimmed. Specular highlights on the nose are near-colourless and jaw shadows are oversaturated; both drag a plain mean away from the real skin colour.
- Hair from a band above the eye line, excluding anything within ΔE 14 of the measured complexion — which keeps forehead out without assuming what colour the hair is.
Skin classification uses the Chai & Ngan (1999) YCbCr bounds combined with the Kovac et al. (2003) RGB rule. The photo itself is never uploaded; only the model and runtime are fetched, and both are cached after first use.