This chapter is about making a composite physically believable by hand: building the image back-to-front in the order of the scene's real depth, and resolving light and color at every layer before the next one goes on top. It covers the depth-ordered build sequence, eyedropper-sampled light and color matching, brush blend-mode targeting for constrained painting, and atmospheric perspective faked by adjusting levels and painting haze — all fully manual, no AI model call involved. The material rests on a single worked example (PHLEARN's sci-fi panorama course), which is both its strength and its main limit.
In most editing, layer order is something you correct after the fact — you notice an element is sitting in front of something it should be behind, and you drag it down the stack. The organizing principle of manual compositing inverts that. Layer order isn't cleanup; it's the plan. You build the image in the same order as the scene's physical depth: the farthest element first, then the next nearest, working layer by layer toward the camera. That's Depth-Ordered Compositing Workflow (Background-to-Foreground Build Sequence).
The rule that makes it work is stricter than "start at the back." Each stage is fully resolved — light, color, integration all finished — before the next, nearer layer is added on top of it. The reason is causal rather than aesthetic: every subsequent layer has to match what's already been established behind it. A foreground figure is lit by the ground and sky it's standing in front of, so those have to be finished before there's anything to match against. If you add the figure first and settle the background later, you've made the figure's light a guess, and you'll be re-matching it every time the background moves.
It's worth separating this from the plumbing question. Photoshop Foundations: Non-Destructive Editing & Selection covers layer structure — how a document stays editable, what lives in a smart object, how selections become masks. Depth ordering is about layer sequence: the order the layers get built in, not just how they're kept re-editable afterward. The two are complements. Non-destructive structure means you can go back and fix the background; depth ordering means you shouldn't have to.
The concrete example the chapter is built on comes from PHLEARN's sci-fi panorama course, and it's worth walking in full because it shows what "resolve each layer, then advance" costs in practice. The whole build is manual — no generative step anywhere in it.
Notice the shape of the sequence: it is depth order almost exactly. Sky and planets, then far mountains, then things on the mountains, then lights, then near-camera creatures and people. The single exception is the finishing pass, which is deliberately global and deliberately last — grade and grain are applied across everything at once, because their job is to erase the evidence that the layers came from different cameras. The texture-unifying end of that idea (matching grain across a composite so mismatched sources stop announcing themselves) is treated more fully in Generative Fill & AI Compositing Toolkit, where the same problem shows up with AI-generated pixels instead of photographed ones.
One honest caveat: this is a single worked example. The chapter's material carries no other documented build, so the seven stages should be read as one course's demonstration of the principle, not as a canonical stage list that every composite follows.
The core integration move in this chapter is small and repeatable: eyedropper colors directly out of the scene's ground and sky, then hand-paint that sampled light onto the cut-out element as if it were reflecting off it. That's Manual Light/Color Matching via Environment Sampling (Non-AI Composite Integration), and it is what turns a pasted subject into a present one. The sampling step matters more than it sounds — the color you paint isn't chosen by eye or by taste, it's taken from the pixels the light in that scene actually produced, so the match is inherited rather than invented.
In the panorama build it's used at three different depths, which is the best evidence that it's a general technique and not a one-off trick. It's what makes near-camera rock photos read as distant mountains, it's what integrates the foreground animal, and it's what integrates the cut-out people. That last case carries a detail worth stealing: the people's original ground was sand, and the sand was matched to the new terrain, so the light plane already lined up before any color work started. In other words, the ground the subject was standing on when photographed is itself a light source bouncing up into them, and matching that first means the color pass has less to fight.
This is the manual counterpart to the AI approach. Photoshop's Harmonize — covered in Generative Fill & AI Compositing Toolkit — chases exactly the same goal, matching a composited element's light and color to its new surroundings, but reaches it through a model call rather than through sampling and brushwork. Same target, two different routes. The reason to know the manual route isn't nostalgia: it's that hand sampling gives you a per-region decision on which part of the scene is lighting which part of the subject, which a single global "harmonize" pass doesn't expose.
The mechanics of applying the sampled color — how you get a stroke to land only on the highlights, or only on hue without dragging luminance with it — are the subject of the next section.
The usual way to paint light onto a composited subject is to paint freely and then mask the damage back — brush over everything, then erase where it shouldn't have gone. Hand Painting via Blend-Mode Targeting (Highlights/Shadow/Color) replaces that with constraint at the source: you set the Brush tool's own blend mode, not the layer's, so each stroke is confined to a specific tonal or color range before it lands. A mode that paints only hue and saturation without touching luminance; a mode that affects only highlights; a mode that affects only shadow.
The distinction between the brush's blend mode and the layer's is the whole point and is easy to skim past. A layer blend mode governs how a finished layer composites against everything under it. A brush blend mode governs how each individual stroke deposits into the pixels it touches. Setting it on the brush means one custom brush can add a highlight here, a shadow there, and a local color shift somewhere else without repainting or re-masking the surrounding area each time — the constraint travels with the stroke instead of being reconstructed afterward.
The technique extends past retouching into outright construction. You can build a bespoke brush shape from scratch rather than reaching for a stock brush, load it with color sampled from the surrounding scene via Manual Light/Color Matching via Environment Sampling (Non-AI Composite Integration), and then stamp it repeatedly to construct complex organic form. In the panorama, the towering spires on the mountains are built this way — entirely by hand, no photography composited in at all — which is stage four of Depth-Ordered Compositing Workflow (Background-to-Foreground Build Sequence). It's a useful reminder that not every element in a believable composite has to have been photographed; it only has to carry the scene's light.
The chapter is thin on specifics here, and it's better to say so than to guess: the material describes the kinds of blend modes used but doesn't name the individual Photoshop modes, and it doesn't describe how the custom brush shape was authored. You can take the principle from this chapter; the exact settings aren't in it.
Distance in a photograph isn't primarily a matter of scale — it's a matter of what the air does to light on its way to the lens. Over kilometers, atmosphere lifts the blacks, drains saturation, and shifts everything toward the color of the sky between you and the object. Atmospheric Perspective Simulation (Distance via Light & Color) is the deliberate faking of that: take a near-camera photographed object, adjust the source photo's light levels, then hand-paint environment-sampled color over it to introduce the haze and desaturation a real telephoto view of a distant object would have.
The panorama's distant mountains are the demonstration. They are ordinary rock photographs — objects that were a few feet from the camera — and what converts them into a horizon line is levels work followed by painted haze in colors pulled from the surrounding sky via Manual Light/Color Matching via Environment Sampling (Non-AI Composite Integration). Once they read as distant, they become the substrate for the next stage, the hand-built spires from Hand Painting via Blend-Mode Targeting (Highlights/Shadow/Color). That ordering is the depth principle in miniature: the mountains had to finish being far away before anything could be built standing on them.
It's worth contrasting the mechanism with the slider-based way of achieving a similar look. Pulling Clarity, Texture, and Dehaze down — the Effects panel work in Lightroom Mobile: Masking, Sliders & Simulated Light — chases the same end state, an element that reads as far away or ethereal, but it does it globally and parametrically. Atmospheric perspective simulation is per-object and painted: you decide which parts of which element get how much haze, and the haze color comes from the actual scene rather than from a slider's idea of it. Same destination, genuinely different mechanism, and the chapter's material doesn't rule on when one is preferable to the other.
Read together, the four techniques in this chapter answer one question: where does physical believability come from when nothing generates it for you? The answer they converge on is that believability is accumulated, not applied. It comes from resolving one depth plane at a time (Depth-Ordered Compositing Workflow (Background-to-Foreground Build Sequence)), from taking the light off the scene rather than inventing it (Manual Light/Color Matching via Environment Sampling (Non-AI Composite Integration)), from strokes that can only affect the tonal range they're supposed to (Hand Painting via Blend-Mode Targeting (Highlights/Shadow/Color)), and from reproducing what distance physically does to an object (Atmospheric Perspective Simulation (Distance via Light & Color)). None of those is a single move. Each is a discipline applied repeatedly.
That's the chapter's argument against treating a one-click tool as the whole answer. The AI compositing stack in Generative Fill & AI Compositing Toolkit is genuinely fast, and Harmonize solves in one call what environment sampling solves by hand — but it solves it as one decision for the whole element. The manual techniques here are the vocabulary for the cases where that single decision isn't enough: an element that needs different light on its upper and lower halves, a distance cue that no slider produces, a form that has to be built rather than found. Knowing them is also what lets you diagnose an AI result — if a Harmonize pass looks wrong, the manual model tells you what is wrong, whether it's the light plane, the haze, or the bounce color off the ground.
A plain note on the limits of what's here. This chapter has no sources of its own beyond the single PHLEARN course example, and that shapes what it can and can't give you. It gives a solid organizing principle and four named techniques with a worked demonstration of each. It does not give named Photoshop blend modes, numeric values, brush construction steps, or any stated criterion for when a match is good enough to move on. Those gaps are real, and the honest way to use this chapter is as a model of how to think about a composite's physics, with the button-level specifics to be filled in from practice or from a source the chapter doesn't currently carry.