Why Your Color Season Looks Different in Every Light

You run a color analysis in your bathroom and it says you are warm. You run it again by the window an hour later and it says you are cool. Nothing about you changed in that hour, so one of those answers is wrong, and possibly both.

This is the single largest source of error in any color analysis done from a photograph, and it is worth understanding properly, because once you know what is happening you can get a reliable reading in about thirty seconds.

Your skin does not have a color. It has a response.

A camera does not record what your skin is. It records the light that bounced off your skin and reached the sensor. Change the light, and you change the measurement, even though your skin is identical.

The property that matters is called the color temperature of the light source, measured in kelvin. Lower numbers are more orange, higher numbers are more blue:

That is a span of more than 6,000 K between a bedside lamp and a shaded doorway. A warm bulb pushes every skin tone toward orange. Open shade pushes every skin tone toward blue. Neither is telling you anything about your undertone.

This is why the bathroom mirror and the changing room mirror disagree, and it has nothing to do with your eyes.

What your phone is already doing about it, and why it is not enough

Your camera runs automatic white balance. It looks at the scene, guesses what the light source is, and shifts the colors to compensate so white objects look white.

It is a guess, and it is a guess based on an assumption: that the average of everything in frame should come out neutral grey. That assumption is called the grey-world hypothesis, and it works well in a normal, mixed scene.

It breaks in exactly the situations you take selfies in:

• A strongly colored wall or backdrop. A red wall makes the average of the scene reddish, so the camera compensates by pulling the whole image toward cyan. Your skin now reads cooler than it is.
• A large area of one color in frame. A bright green sweater does the same thing in the opposite direction.
• Mixed lighting. A window on one side and a warm lamp on the other means there is no single correct white balance. Half your face is one temperature, half is another.
• Very dark or very bright backgrounds. These skew the average badly and the correction overshoots.

So the camera makes a correction, and then that correction is itself wrong, and the error lands directly on the thing you are trying to measure.

Why this hits undertone hardest

Color analysis reads three things about you: how deep your coloring is, how much contrast there is between your features, and whether your undertone leans warm or cool.

Depth and contrast are fairly robust. If your hair is much darker than your skin, that stays true under a warm bulb and under daylight. The relationship survives even when the absolute values shift.

Undertone does not survive. Undertone is a small difference in the balance between the red and yellow components of your skin. In the measurement system most analysis uses, we are talking about a shift of a few units on a scale where a warm bulb can move you fifteen or twenty. The signal is smaller than the noise.

That is why the same person gets "warm autumn" in one photo and "cool summer" in the next. The method is not broken. The lighting overwhelmed it.

This applies to every photo-based color analysis, including ours. Anyone who tells you their tool is immune to lighting is either not measuring undertone properly or not telling you the truth. What a good tool can do is detect when the lighting was poor and say so, rather than handing you a confident answer built on a bad measurement.

How to get a reading you can actually trust

Five things, in order of how much they matter.

1. Face a window during the day. Do not stand with your back to it. North-facing is ideal in the northern hemisphere, south-facing in the southern, because indirect daylight is the most stable light source you have free access to. Let the light fall on your face, not behind you. Backlighting throws the exposure off and puts your face in shadow.

2. Turn off every artificial light in the room. This is the step people skip and it is the one that matters most. Mixed daylight and tungsten is worse than either alone, because no single correction can fix both. One light source, one temperature, one correctable image.

3. Avoid the two extremes of daylight. Direct midday sun blows out highlights and kills the subtle information in your skin. Golden hour, the hour after sunrise and before sunset, is beautiful and completely useless here, because the light itself is around 3,000 K and everyone reads warm. Mid-morning or mid-afternoon, indirect, is the sweet spot.

4. Clear the frame. Neutral wall behind you if you can. No red door, no bright green plant, no vivid sweater. If you have a white or grey top, wear it. You are trying to give the white balance algorithm an easy job.

5. Take off makeup, and push your hair back. Foundation is designed to alter your apparent undertone; that is its entire function. Blush and bronzer do the same. Hair falling across your cheeks bounces its own color onto your skin.

Do those five things and you have removed most of the error. It genuinely takes about thirty seconds.

How to tell if a reading is trustworthy

A few checks you can run yourself:

• Run it twice, in different sessions. Same light, different day. If you get the same season both times, the reading is stable. If it flips, your lighting is not controlled.
• Check whether the tool tells you its confidence. A tool that gives you the same certain answer for a photo taken in a dim bathroom and one taken by a window is not measuring lighting quality at all.
• Be skeptical of a result from a photo you already know was badly lit. If you took it at night under a lamp, the answer is a coin flip regardless of who produced it.

Try it with a controlled photo

Our free color analysis tool runs the same AI engine as the app: your photo is resized in your browser and sent over an encrypted connection to Google's Gemini API for the analysis. Neither we nor our Worker write it to storage. If the service is unreachable, the analysis runs entirely in your browser instead. It corrects for the light source, measures your undertone and contrast, and gives you your season with five colors to start from. When it cannot get a clean read, it tells you rather than guessing.

Take one photo in bad light and one by a window, and run both. Seeing the two answers side by side is the fastest way to understand how much the light was doing.

Common questions