Light Has a Temperature and Your Subject Remembers It
Every photograph records the color of its light source. A product shot near a window at midday carries a cool cast, slightly blue in the highlights. The same product under a warm bulb in a stockroom carries an orange one. Your eye ignores this in the original because the whole frame shares it and your visual system adapts to the average.
Cut the subject out and the adaptation breaks. Now a subject lit at roughly 5500 kelvin sits on a background built around 3000, and the mismatch has nowhere to hide. Nobody computes color temperature by eye. Everybody notices when two halves of one image disagree about where the light came from.
Clean Edges Are the Easy Half
The tooling has become very good at the part that used to be hard. Pixelcut's background remover returns a usable alpha channel on a difficult subject in seconds, on a job that used to be billed by the hour, and on most subjects the result beats a rushed hand-mask. What no automated matte can do is decide what the subject should look like in its new surroundings. That's a color decision rather than a selection decision, and it stayed with you.
The Halo Is Made of the Old Background
Look closely at the boundary of an automated cutout and you'll usually find a rim one or two pixels deep carrying the color of whatever used to sit behind the subject. A bottle photographed against a white wall keeps a faint white outline. The same bottle shot against foliage keeps a green one.
On a white background nobody sees it. Move that cutout onto deep navy and the white rim reads as a glow. This isn't a masking failure. Boundary pixels genuinely contain both things, and the matte has to assign them somewhere.
Shadows Are Not Gray
The most common tell in a composite is a shadow rendered as a soft gray blob. Real shadows aren't gray. A shadow is the region a light source can't reach, which means it's lit by everything else in the scene: the sky, the walls, the bounce off a nearby surface. Outdoors that makes shadows distinctly blue. In a room with warm walls it makes them warm.
So when you add a contact shadow under a cutout, sample its color from the background's own darker regions instead of reaching for black at low opacity. The gap between a gray shadow and a shadow tinted with the scene's ambient color is most of the distance between "obviously composited" and "fine".
Gradients Are Unforgiving Backgrounds
A flat color hides a lot. A gradient hides nothing, because the subject's edge crosses a moving target: the top of the cutout meets one hue and the bottom meets another, so a single edge correction can't be right in both places.
Gradients also carry a directional claim. A background running bright at the top to dark at the bottom implies light from above, and a subject lit from below will argue with it even when every color value matches.
Compositing onto a gradient, then, starts with choosing a gradient direction that agrees with the subject's existing light. That's far cheaper than relighting the subject.
A Five-Minute Check That Catches Most of It
- Blur the whole composite heavily. Mismatched brightness and color survive the blur and become obvious.
- Invert the image. Edge halos and stray fringe pixels are far easier to see in negative.
- Desaturate everything. If the subject still pops out unnaturally, the problem is tonal rather than chromatic.
- View it at thumbnail size. Most composites that fail fail worse when small.
Perceptual Space Is Where Matching Happens
Nudging RGB values until two elements agree is unreliable, because equal steps in RGB don't produce equal perceived differences. Shift red by ten and the change you see isn't the same size as shifting green by ten. Which is why color matching that has to hold up gets done in a perceptual space, and why the vocabulary comes from colorimetry rather than from image editors. The International Commission on Illumination defined the CIELAB space for this exact purpose in 1976, then later recommended the CIEDE2000 difference formula to correct the regions where CIELAB itself turned out not to be perceptually even.
In practice that means judging a match on lightness first and the two opponent axes second, rather than eyeballing RGB sliders. Most editors will show you Lab values if you ask.
When to Stop Matching and Start Choosing
There's a point where matching becomes the wrong goal. If the subject was shot under light that can't plausibly coexist with the background you had in mind, the honest move is to stop correcting and pick a different background. A brand gradient is a preference. The light in the original capture is a fact.
Match the Background to the Subject, Not the Reverse
The subject is usually the expensive, unrepeatable half of the image and the background is the cheap, infinitely adjustable half. Designers instinctively work the other way around, treating the background as fixed and grinding away at the subject to fit it. Reverse that order and most of the labor disappears. Pick a background whose implied light matches the light your subject was already carrying, and the composite stops announcing itself.