Indoor light and the colours it invents

A white wall is never white. It is whatever colour the light landing on it happens to be, and indoors that is almost never daylight — which is why so many photographs of ordinary rooms come out looking like they were taken through weak tea.

Every light source has a colour

Colour temperature is a way of describing that colour with one number, borrowed from what happens when you heat a piece of metal: warm at first, then yellow, then white, then blue. Counter-intuitively the numbers run the other way round from the words — a candle is about 1,900 K and reads as very warm, midday daylight is about 6,500 K and reads as neutral, and an overcast sky can pass 8,000 K and reads as cold.

Common sources, and how far each sits from daylight
SourceApprox. temperatureFrom 6,500 K
Candle1,900 K−4,600
Incandescent bulb2,700 K−3,800
Warm white LED3,000 K−3,500
Warm fluorescent tube3,500 K−3,000
Cool white tube4,200 K−2,300
Overcast daylight7,000 K+500

Your eye discounts all of this without being asked. Walk from a street into a kitchen and for a second everything looks amber, then it stops — the same adjustment that lets you read white paper as white under any of the sources above. A camera has to perform that adjustment as a calculation, once, for the whole frame, and it gets it wrong indoors more often than anywhere else.

Why the guess goes wrong indoors

Automatic white balance works by finding something in the frame it can assume is neutral and asking what colour it came out. Outdoors that assumption usually holds: there is sky, there is concrete, there is a broad spread of colours that average to something close to grey. A small room is the opposite kind of scene. It might be three quarters wooden furniture, or a single painted wall, or a tablecloth — and any of those makes the average of the frame genuinely not neutral, which sends the whole estimate off in the direction of the dominant object.

Warm LEDs are a recent complication

A great deal of indoor lighting has been replaced in the last decade, and much of the replacement was chosen to look like the filament bulbs it displaced. Those LEDs sit around 2,700 to 3,000 K and, unlike a filament, produce their light from a phosphor with an uneven spectrum. Two rooms can measure the same temperature and photograph noticeably differently, which is why a single correction factor per room is the best any estimate can do.

Tube light is worse than its number suggests

Fluorescent tubes emit spikes rather than a continuum, with a strong one in the green. That is off the warm-to-cool axis entirely, so it survives a temperature correction untouched. If a photograph still looks slightly unwell after the warmth is right, the residue is almost always green, and the Tint control is the one that addresses it.

When not to correct

The most common mistake with white balance is applying it to a scene whose colour was the point. A room lit by one lamp is warm, and that warmth is a large part of what the photograph is about; neutralise it and you get an accurate record of a place that now looks like an office. The tool weakens its own correction when its two estimates disagree, which catches sunsets and strongly coloured rooms most of the time — but it is a heuristic, and you are the one who was there.

The practical approach is to look at the kelvin figure the panel prints. If it is somewhere between 800 and 2,000 below daylight on an indoor photograph, it has probably found a real cast and removing it will give you back the wall colour you remember. If it is much larger than that, look at the picture before you accept it: a big number usually means the frame was dominated by one strongly coloured thing, and Strength is the fastest way to say “less than that”.

The questions that come up

Why do photographs taken under tube lights look green as well as yellow?
Because a fluorescent tube does not emit a smooth spread of colours the way a filament or the sun does — it emits a few narrow spikes, and one of the strongest sits in the green part of the spectrum. Colour temperature only describes where a light sits on the warm-to-cool axis, so a tube can be corrected to the right temperature and still be visibly green. That is what the Tint control is for, and it is why it exists separately from Warmth rather than being folded into it.
My room has a window and a lamp. Which one does the tool correct for?
Neither, exactly. It estimates one illuminant for the whole frame, so with daylight on one side and a warm lamp on the other it lands somewhere between them and leaves each side slightly wrong in opposite directions. There is no global fix for mixed lighting, because the correct answer differs from one part of the picture to another and this page deliberately has no way to treat one part differently from another. The practical move is to decide which half of the frame matters and use Warmth to favour it.
Should I correct the cast at all?
Often not. Warm light at the end of a day, a room lit by a single lamp, candles on a table — in all of those the colour is the subject rather than an error, and neutralising it produces a technically accurate photograph of a scene nobody remembers. A useful test is whether you would have described the light to someone. If you would have said the room was warm, keep it warm, and use Strength rather than Warmth so the exposure and contrast work still happens.
Why does the correction sometimes barely move a very yellow photograph?
Because two estimates disagreed. The pass compares the average colour of the frame against the colour of its brightest few percent, and where those two answers diverge it treats the divergence as evidence that the scene is genuinely coloured rather than miscast. A photograph of a sandstone wall at sunset is very yellow and correctly so, and it produces exactly that divergence. The safeguard costs you some correction on unusual indoor scenes and saves you from having every sunset drained.

Take it back to the picture

Corrected on your own machine — we never get a copy.