Snagfit records five things about every garment it finds, and one of them is the colour. That field is built from what the photograph contains, not from what the garment is in daylight.
Which means light, rather than the garment, often decides what the search goes looking for.
What does bad light do to a search?
A navy coat photographed under warm indoor light reads as brown. Snagfit describes a brown coat, because that is what is in the picture, and the shopping search returns brown coats.
Nothing malfunctioned. The description is an accurate account of the image. The image is an inaccurate account of the coat.
This is the most common reason results come back in a colour that feels wrong, and it is invisible until you know to look for it.
In what four ways does light change a garment?
Colour temperature. Household bulbs are warm and push everything towards orange. Some office lighting is cool and pushes towards blue. Both shift a garment's apparent colour by enough to change the word used for it.
Colour grading. Video is graded for mood. A scene pushed blue for cold or amber for warmth carries every garment in it along.
Filters. A filter applied to a post changes colour and usually contrast as well. Whatever it did to the skin tones it also did to the clothes.
Shadow. Not a colour shift, but a loss. Detail in shadow is simply not recorded, so the texture and seams of a garment on the dark side of a person do not exist in the file.
Why are dark clothes the hardest?
Outerwear is disproportionately black, navy and charcoal, and dark garments are where this hurts hardest.
In good light, a black wool coat shows a weave, a seam line down the front, the buttons, the way the lapel folds and the point where the hem falls. All of that goes into the description.
In low light, the same coat is a shape. There is no weave, no visible closure, no lapel detail. What Snagfit can honestly say is "black coat", and that matches an enormous number of listings.
So the frustration of "it just returns generic black coats" is usually a lighting problem rather than a search problem, and the fix is a brighter frame rather than another search.
What is the best light to photograph an outfit in?
Even daylight, ideally falling on the front of the person.
Daylight renders colour closest to true, which is why shops photograph products under lights chosen to imitate it. Front lighting means the side of the garment facing the camera is the side that is lit, so texture and seams are visible rather than hidden.
Overcast daylight is better than bright direct sun. Direct sun creates hard shadows, which means half the garment is bright and half is black, and the dark half contributes nothing to the description.
Indoors, near a window is much better than under a ceiling light. Ceiling light comes from above, which shadows everything that faces forward and makes footwear particularly bad.
Does brightening a dark photo help?
A tempting fix that makes things worse.
Raising the brightness of an underexposed image amplifies whatever was recorded, including noise. Colour that was never captured cannot be recovered by turning the exposure up. What you get is a lighter version of the same missing information, plus grain.
Snagfit then describes a noisy, washed image, which is not an improvement on describing a dark one.
The genuine fix is a different frame. If the outfit appears anywhere else in better light, that frame is worth thirty seconds of scrubbing.
Why are film and video the hardest?
Two reasons, and they compound.
Productions light for mood rather than for accuracy. Night scenes are graded blue, candlelit interiors are pushed orange, and moody dramas are frequently dark on purpose. That is the look they wanted, and it makes every garment in the frame unreliable in colour.
Video compression then removes fine texture, particularly in dark areas, which is exactly where a dark garment's detail would have been.
The fix is the same: find a brighter scene showing the same costume. Most productions light at least some scenes conventionally, and one daylight shot is worth more than any amount of work on a night scene. There is a post on screen costumes specifically.
Which colours go wrong most often?
Some colours are far more fragile under bad light than others, and knowing which tells you when to be suspicious of a result.
Navy. The single worst offender. Under warm light it reads as brown or black. Under cool light it reads as a brighter blue than it is.
Charcoal and dark grey. Frequently collapse into black in any shadow.
Camel, tan and beige. These sit close together to begin with, and a warm or cool cast slides a garment from one to the next.
Olive and khaki. Shift towards brown in warm light and towards grey in cool light.
Burgundy and maroon. Read as brown in low light and as red in bright light.
Off white, cream and ivory. Any warm light turns all three into the same colour.
Bright, saturated colours hold up much better. A true red or a strong emerald survives poor lighting far more reliably, because there is more distance between them and the nearest alternative.
So if the garment you want is navy, charcoal or a neutral, treat the lighting question as the main one. If it is a strong colour, you have more room.
How can skin tone tell you if the light is off?
A practical trick for judging a frame before you upload it.
Look at the person's face and hands. If the skin looks orange, the whole image is warm and every garment in it has been pulled towards yellow and brown. If the skin looks blue or grey, the image is cool and colours have gone the other way.
Skin is something everyone can judge instinctively, whereas judging whether a coat is the right shade of navy is much harder without a reference.
When the skin looks natural, the garment colours in that frame are probably close to true, and the description you get back will be reliable. When it does not, expect the colour word to be off and plan to correct it by hand.
What does lighting not affect?
Worth stating, because people worry about the wrong things.
Screenshot brightness. A screenshot captures image data, not what your eyes see, so taking one while your screen is dim does not produce a dark file. The exception is photographing a screen with another camera, which does inherit the lower brightness along with reflections.
Resolution. Every upload is downscaled to 1280 pixels on its longest side anyway, so a well-lit modest photo beats a badly lit high resolution one every time.
Interface elements. Captions, buttons and progress bars are ignored, because Snagfit is looking for garments.
What happens when a photo has two kinds of light?
One case that catches people out because the photo looks fine.
A person standing near a window with a warm lamp behind them is lit by two different colours at once. One side of the garment reads warm, the other cool, and neither is the true colour.
Snagfit describes what is dominant, which may be either. If a result comes back in a colour that does not match either side of what you can see, mixed lighting is usually the reason, and a different frame is the fix.
How do you pick the best frame?
A short checklist that takes about ten seconds.
Is the garment lit from the front rather than behind? Backlighting turns a person into a silhouette and removes everything.
Can you see texture in the darkest part of the garment? If the shadows are solid black, detail is missing there.
Does the colour look approximately right to you? If the whole image is obviously orange or blue, expect the description to follow.
Are the feet lit? Footwear is almost always the darkest part of a photo, so if you want the shoes, judge the frame on them rather than on the overall exposure. There is a post on shoes covering why.
What if there is only one photo and the light is bad?
Sometimes there is only one photo and the light is what it is.
Search it anyway, then read the returned colour with scepticism. If Snagfit says brown and you remember navy, search the garment name with the colour word swapped. The rest of the description, the cut, the closure, the material, is still doing useful work.
This is one of the cases where taking the garment name somewhere else pays off. The words transfer, and you can correct the colour by hand in a way the original search could not.