Render Passes & Compositing

When an engine computes a final image, it is internally calculating many separate things at once - how much colour each surface has, how much light bounces onto it, how strongly it reflects, how deep the contact shadows sit in its crevices, how far each pixel is from the camera. In a normal single-image render, all of that gets flattened together into one result. The information is still there in the calculation; the engine just throws away the breakdown and hands you the sum.

A multi-pass render keeps the breakdown. Instead of one flattened picture, the engine writes out each of those components as its own image - a diffuse pass, a reflection pass, an ambient-occlusion pass, and so on. Recombine them by simple addition and you get back exactly the original beauty image; but now each ingredient is a separate layer you can turn up, turn down, tint or mask independently. The reflection too hot on the glass? Lower the opacity of the reflection layer. Contact shadows too weak to read? Multiply the AO pass a little harder. None of this touches the 3D scene, and none of it needs a re-render.

This matters enormously for a simple economic reason: a re-render can cost minutes or hours; a composite adjustment costs seconds. The single most expensive thing in visualization is your render time, and passes let you make ninety percent of your final look-decisions after that expensive step is already done. It also protects you from a very common trap - endlessly re-rendering to chase a look you could simply have graded. Output your passes and you buy yourself a huge, cheap creative playground on top of one render.

The passes are usually written as a high-bit-depth format such as EXR, which stores light values in the full high dynamic range the engine computed rather than crushing them to screen brightness. That extra range is what lets you recover a blown highlight or lift a dark corner in the composite without the image falling apart - the data is still there to pull from.

Engines can output dozens of passes, but a working archviz composite leans on a familiar handful. The diffuse (or albedo) pass is the raw surface colour with no lighting - flat, matte, the pure material colour of every object. The lighting / GI passes carry the direct and indirect illumination separately, so you can rebalance sun against bounce. The reflection and specular passes hold everything shiny - the sky in the glass, the sheen on a polished floor, the glint on metal - isolated so you can dial reflections up for drama or down when they overwhelm.

The ambient occlusion (AO) pass is a greyscale map of contact shadow: it darkens where surfaces meet, where objects touch the floor, inside tight crevices. Multiplied gently over the image it re-grounds everything and adds the subtle depth that separates a floating, pasted-on look from a solid one. The z-depth pass encodes distance from the camera as brightness - near is white, far is black, or the reverse - which is pure gold in post: it lets you add realistic distance fog, a graded atmosphere, or a believable depth-of-field blur that falls off with distance, all controlled by that one greyscale map.

Finally, masks - object IDs, material IDs, or the modern favourite cryptomatte - give you pixel-perfect selections of any object or material with a single click. Want to recolour just the brick, brighten only the timber, or select every window at once? A mask pass hands you that selection instantly, instead of you painstakingly cutting it out by hand. Together these passes cover the vast majority of what any archviz composite needs: colour, light, reflection, contact shadow, depth and selections. Learn to read each one and you can diagnose and fix almost anything without leaving Photoshop.

Compositing is the craft of stacking those passes back together and steering the result. In a still-image workflow that usually means Photoshop; for animation and moving sequences it means After Effects (or a node compositor). The mechanics are the same idea in both: each pass becomes a layer, and the blend mode of each layer decides how it combines with the ones below. Reflection and specular passes are typically added with a Screen or Linear Dodge (Add) mode, because light is additive. The AO pass sits on Multiply, because occlusion darkens. The diffuse forms the base. Stack them in the right order with the right modes and the original render rebuilds itself - but now every layer has an opacity slider and can carry its own adjustments.

That is where the control comes from. On top of the reassembled beauty you add adjustment layers - curves, levels, colour balance, hue and saturation - that grade the whole image or, masked, just part of it. Because the z-depth pass is available, you can load it as a mask and apply an effect that strengthens with distance: a cool haze that thickens toward the horizon, a warm glow that only touches the far facade, a blur that softens the background while the foreground stays sharp. Because you have object masks, you can push one material without disturbing the rest.

The payoff is a kind of freedom the engine alone never gives you. A single render becomes the raw material for many finished images: a bright cheerful daytime version and a moody dusk version from the same passes, a hero shot with strong reflections and a calmer study with them pulled back. You are no longer at the mercy of one baked result - you are directing the final picture, ingredient by ingredient, in real time.

The discipline that holds all of this together is a single principle: never destroy the original pixels. Every serious post-production workflow is built to be reversible, so that any decision made an hour ago can be undone or re-tuned without starting over. In practice this means a few concrete habits. Keep the base render and every pass on their own layers, untouched. Do all your colour and tonal work through adjustment layers rather than by directly altering the render - a Curves adjustment layer grades the image but leaves the pixels beneath it pristine. Confine every effect to a mask so you can paint its influence in and out. And convert imported elements to smart objects so a scaled or filtered layer can be re-scaled or re-filtered later with no loss.

Why such care? Because look development is iterative and taste changes. The client asks for it warmer, then decides the first version was better; a printed proof reveals the shadows are too heavy; you return to the image a week later with fresher eyes. A destructive edit - flattening the passes, painting straight onto the render, baking in a filter - throws away your ability to respond. A non-destructive stack lets you say yes to any of those requests in seconds. It is the difference between a document you can keep negotiating with and a photograph you have already burned.

There is a practical filing habit too: save the layered master (the PSD or AE project) alongside the flattened export, and keep the passes archived. A well-organised, non-destructive composite is not just tidy - it is what lets a render keep earning its keep long after the engine has moved on. Master this workflow and you gain the calm confidence that nothing you do in post is ever truly final until you decide it is.