Last year I pulled up an old composite I’d made during my agency days and genuinely cringed. The subject was cut out cleanly, the background was dramatic, the lighting was dramatic. Everything was technically correct. And it looked completely, obviously fake.

The problem wasn’t the selection. It wasn’t the masking. It was that I had treated blending as the last 10% of the job when it’s actually closer to 60% of what makes a composite believable. I see the same mistake in almost every beginner manipulation I review. The pieces are there. They just don’t belong to the same world yet.

Here’s how to fix that.

The Real Reason Composites Look “Off”

Your eyes are extraordinarily good at detecting inconsistency, even when your brain can’t name what’s wrong. When a subject photographed under flat studio lighting gets dropped onto a background shot in golden hour, something feels wrong before you can articulate it. That feeling comes from three mismatches happening simultaneously: luminosity (how bright each element is relative to its light source), color temperature (the actual warmth or coolness of the ambient light), and edge quality (how hard or soft the transition between elements is given the depth of field in the scene).

Most people fix one of these. You need to fix all three, in order, or the fix you made first will look wrong again after you fix the third.

Match Luminosity Before You Touch Color

Open both images, your subject and your background, and go to Window > Histogram. You want to compare the average tonal range of the two. If your subject’s histogram peaks around 160 and your background peaks around 110, your subject is living in a brighter world than the scene you’re dropping it into.

Add a Curves adjustment layer clipped to your subject layer. Pull the midpoint of the curve down until the histogram peaks align roughly between the two images. Don’t aim for pixel-perfect matching. Aim for the same neighborhood. A subject that’s 8-12 points off in luminosity reads as plausible. One that’s 40 points off reads as pasted.

Once your luminosity is in range, switch your Curves layer blending mode to Luminosity. This locks the adjustment to brightness only and prevents it from contaminating your color in the next step.

Color Temperature Matching with a 1% Sample

Create a new layer above your subject, set the blending mode to Color, and fill it with a neutral gray (128, 128, 128). Now use the Eyedropper tool set to a 5-by-5 pixel average sample and click the brightest neutral area in your background, somewhere like a cloud edge, a concrete wall, or a highlight on the ground. Note the RGB values in the Info panel.

If the background reads as R:255 G:240 B:210, that’s a warm golden hour scene. Your fill layer needs to shift toward those values. Go to Image > Adjustments > Color Balance and push your highlights toward yellow and red until the layer’s tone matches what you sampled. Reduce the layer opacity to between 15% and 35%. Too much and your subject looks tinted. The right amount makes them feel like they’re standing in the same light as everything else.

This two-minute step does more for believability than any plugin I’ve ever bought.

Edge Refinement Is Not About Selection Quality

Here’s the part that took me years to understand. The issue isn’t whether your selection is accurate. The issue is whether your edges match the physics of the scene.

A subject shot with a 50mm lens at f/1.8 will have slightly soft edges on hair and fabric. A background shot with a wide lens at f/11 will be sharp from front to back. Drop a soft-edged subject on a hyper-sharp background and it floats, even with a perfect mask.

Go to your layer mask, apply a Gaussian Blur of 0.3 to 0.8 pixels to the entire mask (Filter > Blur > Gaussian Blur while the mask is selected). That’s it. Sub-pixel softness. Not enough to see as blur, but enough to remove the “cut out” look that comes from a mask with mathematically hard edges that no real camera lens actually produces.

For hair and fine detail, use the Smudge tool at 15% strength directly on the mask and pull individual strands outward two or three pixels. It takes five minutes and it’s the difference between a hair mask that looks extracted and one that looks grown.

The Test That Catches Every Remaining Problem

When I think a composite is done, I do one final check I picked up from a painful project years ago, one where I spent the better part of two days building a single image and only caught a major lighting error in the last hour. I desaturate the entire merged composite (Shift + Ctrl + U on a flattened copy) and look at it in grayscale for 60 seconds.

Color is distracting. Without it, your eye immediately finds the elements that don’t share the same light. Shadows that fall the wrong direction. Highlights that are too bright or too dim. A rim light on your subject that doesn’t match the backlight in the scene. Fix what you find in grayscale and then bring the color back. The image will hold together in a way that’s hard to achieve if you work in color the whole time.

The most important thing I can tell you about photo manipulation is this: realism isn’t about how good your cutout is, it’s about how thoroughly you’ve convinced the viewer that every element was photographed in the same moment under the same light. Get the physics right and the viewer’s brain fills in the rest.