Compositing 3D Renders: What Actually Works

Compositing 3D renders is the stage where separately rendered layers, called passes, are assembled into the finished image. Instead of asking the renderer to produce one flattened picture and living with it, the artist renders diffuse light, specular reflections, shadows, ambient occlusion, depth and object identifiers as distinct channels, then rebuilds the image in a compositing application where each element can be adjusted on its own. A label that reads slightly too warm, a reflection that competes with the logo, a background that needs more atmosphere: in a well-built composite, each of those is a small, localized adjustment rather than a trip back to the render farm.

This handbook is written for the people who commission that work: brand and marketing managers ordering product imagery, agencies subcontracting CGI, architects and developers buying visualization, and in-house teams deciding whether to build the capability or hire it. It covers what to put in the brief, what drives cost, how to compare suppliers who describe the same job in very different ways, the red flags that predict expensive revisions, and how to judge what comes back. The practitioners reading along will find the technical reasoning behind each recommendation, because the buying decisions only make sense once you understand why the pipeline works the way it does.

The central idea is simple. Compositing is where most of the final look is decided, and rendering to passes is what makes adjustment possible without re-rendering. A supplier who plans for that from the start will be faster and cheaper to work with over the life of a project, even if their first quote is not the lowest.

What you are actually buying when you commission compositing

When a studio quotes for 3D imagery, compositing is often buried inside a single line called "rendering" or "post-production." It is worth pulling it apart, because the value you get depends on three things that are easy to overlook in a proposal.

A set of passes, not just a picture

The first thing you are buying is the render output itself, and specifically its structure. A render delivered as a single beauty image can be color-corrected and retouched, but only globally or with hand-painted masks. A render delivered as a multi-channel OpenEXR file carrying the passes the composite needs can be taken apart and reassembled. The difference shows up the first time a stakeholder asks for a change. Our guide to choosing OpenEXR render passes that pay for themselves covers the file format side in more depth; for buying purposes, the point is that the pass list is a deliverable and belongs in the contract.

A composite that someone else can open

The second thing you are buying is the composite itself: the node graph or layer stack in Nuke, Fusion, After Effects, or Photoshop that turns passes into a finished frame. This file encodes every creative decision made after rendering. If it is organized, labeled and handed over, your team or another supplier can make changes months later. If it is a tangle of unnamed nodes that only the original artist understands, you have bought an image and a dependency.

A predictable revision cost

The third thing, and the one that most affects total spend, is how cheaply the work can be changed. A supplier who renders only a beauty pass may quote lower, then charge for full re-renders every time a color, reflection or shadow density needs to move. A supplier who renders to passes and keeps the composite structured absorbs most revisions in the composite. When you compare proposals, you are really comparing expected total cost across the revision cycle, not the price of the first delivery.

Specifying the passes: what to put in the brief

You do not need to be a lighting technical director to write a good brief, but you should state the outcome you want and ask the supplier to name the passes that deliver it. The reference set most studios work from is well established: diffuse, specular, shadow, ambient occlusion, depth, and object and material identifiers. Renderer documentation such as Pixar's RenderMan documentation on arbitrary output variables shows how these are defined at the render stage, and every major production renderer offers an equivalent system.

Map each likely change to a pass

The practical way to build the list is backward from the changes you expect. Think about the reviews your images will go through and the comments that typically come up, then ask which pass makes each comment cheap to address.

• "The product looks flat" or "too contrasty" calls for separate diffuse and specular passes, so shading and highlights can be balanced independently.
• "Tone down the reflection on the logo" calls for specular split by object, or a material identifier that isolates the logo surface.
• "The shadow is too heavy under the bottle" calls for a shadow pass or a separate contact shadow, so density can be adjusted without relighting.
• "It needs more depth" or "soften the background" calls for a depth pass that drives depth of field or haze in the composite.
• "Can we try the cap in navy?" calls for object and material identifier passes, which let an artist select the cap cleanly and shift its hue.
• "Warm up the key light, keep the fill cool" calls for light groups, where contributions from each light, or group of lights, are rendered to their own channels.

Why a beauty-only render is a trap

Rendering only a beauty pass is a frequent and expensive mistake. It looks efficient because the files are small and the render settings are simple, but it moves every later decision back to the renderer. On a still image of modest complexity that may mean an hour or two of machine time per change; on an animation it can mean days. It also forces retouchers to paint masks by hand, which is slow, inconsistent across a series, and fragile at antialiased edges. If a supplier proposes beauty-only output for anything that will go through more than one round of review, ask why.

Do not ask for every pass either

The opposite error is to demand every pass the renderer can produce. Each channel adds render time, disk space and transfer time, and an overloaded EXR slows the composite down. A good supplier will propose a focused set tied to the likely changes and explain what they left out. That explanation is itself a signal of competence. For more on how the files are named, versioned and stored once they exist, see our guide to getting render output management right.

Linear workflow and color management: the question that separates suppliers

If you ask only one technical question during supplier selection, ask how they handle color in the composite. The answer tells you more about the quality of their pipeline than any portfolio.

Why compositing happens in linear light

Renderers calculate light physically: double the light, double the pixel value. Passes are therefore additive. Diffuse plus specular plus the other light contributions rebuilds the beauty image, but only if the addition happens in linear values. Monitors and web images, by contrast, use a display encoding such as sRGB that compresses values to match human perception. If an artist converts passes to display space first and then adds them together, the math no longer matches the light, and the result shows up as muddy midtones, harsh highlight transitions, and color corrections that behave unpredictably from one image to the next.

The correct approach is to keep all compositing operations in linear light and apply the view transform, the conversion to what the monitor shows, for display only. The Academy Software Foundation hosts both OpenEXR and OpenColorIO, the open standards most studios use to store linear high dynamic range images and to manage the transforms between working space and display. A supplier who can tell you which working color space they use, which view transform they apply for review, and how they produce the final output files has a controlled pipeline. A supplier who answers "we just use sRGB" is probably compositing in display space.

What that means for your deliverables

For a buyer, linear workflow has two practical consequences. First, the working files you receive, the EXRs and the composite, will look dark or flat if opened in an ordinary image viewer. That is expected, not a fault, and your supplier should tell you which viewer and transform to use. Second, the final delivery files must be explicitly converted to the target display space, typically sRGB for web and e-commerce, or whatever your print partner or broadcast specification requires. Put the target output space in the brief so there is no ambiguity.

Premultiplication and alpha

A related technical point shows up on product and packaging work. Rendered images with transparency are normally stored premultiplied, meaning color values have already been multiplied by alpha. Color corrections applied to a premultiplied image without first dividing out the alpha produce dark or bright fringes on antialiased edges. Competent compositors unpremultiply before grading and premultiply again afterward. If you have ever received cutouts with a faint halo, this is often the reason. Our article on delivering transparent backgrounds from 3D renders properly covers the edge-quality side of this in detail.

What drives the cost of compositing 3D renders

Pricing for compositing varies widely by region, studio size and the kind of work, so there is no honest single number to quote. What can be stated is which factors push the effort up or down, and those are consistent across suppliers. When two quotes differ significantly, one or more of these is usually the explanation.

Image count and variation

A series of images that share a camera setup, lighting and background can share a composite template. The first image carries most of the setup effort; subsequent ones reuse the structure and need only per-image adjustment. A series where every image has a different environment, product and light rig needs close to a fresh composite each time. Ask the supplier how they will template the work across the series, because that single decision often determines whether the per-image cost falls after the first few images or stays flat.

Number and type of passes

More passes mean more render time, more storage, and more nodes to maintain in the composite. Identifier passes are cheap relative to their value. Light groups can be expensive on complex scenes, because each light's contribution must be written separately. Depth and position passes should be stored at full 32-bit float precision to avoid banding in depth effects, which makes them larger than color passes stored in 16-bit half float.

Resolution and format

Print and large-format outdoor work at high resolution produce EXR files measured in hundreds of megabytes or more per frame once many passes are included. That affects render time, storage, transfer and the responsiveness of the composite, all of which show up in hours. Web-only delivery at typical e-commerce resolutions is much lighter.

Integration with photography or live action

Compositing CG onto a clean studio background is one level of effort. Integrating renders into photographed plates or live-action footage is another: it needs camera matching, lighting matched to the plate, often an HDRI captured on set, shadow catchers, grain matching and careful edge work. If your project involves real photography, expect a separate line for plate integration and ask who owns the camera and lighting match. Our guide to HDRI and image-based lighting explains why on-set capture matters so much for this kind of work.

Revision rounds and approval structure

The number of review rounds included, and how many stakeholders are commenting, often matters more than any technical factor. A composite that has to satisfy a brand team, a legal reviewer and a retailer's specification will go through more iterations than one approved by a single art director. Be honest about your approval chain in the brief.

Handover and archiving

Delivering a clean, documented composite file and the full set of EXRs takes time to prepare. Some suppliers include it; others charge for it or do not offer it at all. If you expect to update the images later, for a packaging change or a seasonal variant, handover is worth paying for.

Questions to ask a compositing supplier before you sign

A supplier's answers to a handful of specific questions reveal more than their showreel. Showreels show finished images; they do not show whether those images were cheap to change, whether the composite was maintainable, or whether the color pipeline was controlled. Ask these questions in writing, and pay attention to how specific the answers are. The broader process of shortlisting is covered in our guide to choosing a 3D studio; the list below focuses on compositing.

Good answers name specific passes, formats and applications, and they draw a clear line between composite changes and re-renders. Vague answers ("we'll render whatever's needed," "our artists handle color") are not necessarily a sign of poor work, but they do mean the risk sits with you. If a supplier bristles at being asked, that is also information.

Comparing compositing quotes line by line

Quotes for the same brief can differ by a wide margin, and the cheapest is not always the one that costs least by the end of the project. The way to compare them is to normalize what each one includes, then estimate the cost of the revisions you are likely to need under each approach. A scorecard makes this concrete and keeps the discussion with stakeholders focused on scope rather than headline price.

Score each supplier from one to five on each row, multiply by a weight (for example three for high, two for medium, one for low) and total the result. Then set that total alongside the quoted price and your revision estimate. A supplier who scores well on the high-weight rows and quotes slightly above the others is frequently the lower-risk choice.

Reading between the lines of a quote

A few patterns recur. A quote that lists "rendering" and "retouching" but never mentions compositing or passes usually indicates beauty-only output with Photoshop cleanup. A quote with a low base price and a high per-revision rate is pricing on the expectation that changes will need re-rendering. A quote that includes a test or look-development stage on a single hero image before rolling out the series is usually from a studio that has learned where projects go wrong; our article on render tests and wedges explains what those early tests should resolve.

Red flags in proposals, files and first deliverables

Most compositing problems are visible early if you know where to look. The warnings below come from the mistakes that reliably drive up cost: beauty-only rendering, compositing in display space, re-rendering for changes a composite would have covered, and composites nobody but the author can follow.

The pattern to watch is the ratio between the change requested and the time it takes. A hue shift on a single material should be minutes of composite work once a material identifier exists. If it comes back the next day with a render invoice, the pipeline was not built for revision.

Consistency across a set is one of the main reasons to commission CGI rather than photography in the first place. When it fails, the cause is almost always in the color pipeline rather than the artist's eye, and it will keep recurring until the pipeline is fixed.

Smaller warning signs

• Halos or dark fringes around cutouts, usually from grading premultiplied images without unpremultiplying.
• Stair-stepped or noisy edges in depth-of-field blur, often from depth passes stored at low precision or filtered in a way that breaks at edges.
• Banding in soft gradients, which can come from 8-bit intermediate files somewhere in the chain.
• Different noise character in different parts of the frame, which can happen when passes are denoised individually with inconsistent settings and then recombined.
• Missing contact shadows where a product meets a surface, a sign that the shadow pass or occlusion was not rendered or not used.

A worked example: a twelve-image product series

The numbers below are illustrative. They are chosen to show how the choice of pipeline affects the cost of revisions, not to represent any specific project or any supplier's pricing. Adjust them to your own render times and rates.

Imagine a brand commissioning twelve still images of a cosmetics bottle for e-commerce and social use, delivered at 4000 by 4000 pixels. The bottles sit on a shared rendered set with the same camera height and three-light setup across the series, with variations in angle and props. Assume each final-quality frame takes about 45 minutes of machine time to render in its full configuration.

Pipeline 1: beauty-only rendering

The supplier renders twelve beauty images and retouches them in Photoshop. The first delivery is fast. In the first review, the brand asks for three changes: the cap should shift from charcoal to navy on all twelve images; the reflection on the front label should be reduced because it obscures the product name on four images; and the background should be softer to make the product stand out on six images.

Without identifier passes, a clean cap color change on antialiased and reflective surfaces is hard to paint convincingly, so the supplier changes the material and re-renders all twelve images: roughly 9 hours of machine time. The label reflection is a lighting change, so four images are re-rendered with an adjusted light: about 3 hours. The softer background is achieved with render-time depth of field on six images, which typically increases render time; call it 6 hours at a conservative estimate. That is roughly 18 hours of machine time for one review round, plus artist time to set up each change, check the renders and re-retouch. A second round of feedback would repeat much of this.

Pipeline 2: passes and a structured composite

The supplier renders each frame to a multi-channel EXR with diffuse, specular, shadow, ambient occlusion, depth, and object and material identifiers. Assume the extra channels add around 10 to 15 percent to render time, depending on the renderer and scene, so the initial render grows from 9 to roughly 10 hours for the series. The supplier builds one composite template and applies it to all twelve frames.

The same three review comments are handled in the composite. The cap color change uses the material identifier to isolate the cap and shift its hue; once built in the template, it propagates to all twelve images. The label reflection is reduced by lowering the specular contribution within the label's identifier matte on four images. The background softening uses the depth pass to drive a defocus in the composite on six images. None of these changes needs a new render, and the artist time involved is measured in hours rather than days.

File size arithmetic

The larger files in pipeline 2 are real and worth planning for. At 4000 by 4000 pixels, one 16-bit half-float channel is 16 million pixels at 2 bytes each, or about 32 MB uncompressed. An RGBA beauty image is four channels, about 128 MB uncompressed. If the EXR carries, say, 20 channels of color passes in half float plus a depth channel in 32-bit float, the uncompressed size approaches 700 MB per frame. Lossless compression options such as ZIP or PIZ typically reduce that considerably, and lossy options such as DWAA reduce it further for color channels, though identifier and depth data should stay lossless. Across twelve frames and a couple of versions, you are still likely to be managing tens of gigabytes. That is ordinary for production work, but it should be in the plan and the transfer method agreed in advance.

The conclusion of the example is not that passes are always cheaper. On a single image approved in one round, the beauty-only pipeline may genuinely cost less. The passes pipeline wins as soon as there is a series, more than one review round, or any likelihood of future variants, which describes most commercial product work.

Judging the delivered composite

Once work comes back, you need a way to review it that goes beyond "does it look good." A structured review catches the problems that surface later, when a variant is needed or the images are used at a larger size. Our general guide to reviewing 3D work covers feedback process; the checks below are specific to compositing.

Review the image at the right size and in the right space

Look at finals at 100 percent on a calibrated display, in the output color space you specified. Many problems, including edge halos, banding and noisy depth-of-field transitions, are invisible at fit-to-screen size. If the images will be used in print, review a proof from the print partner, not a screen export.

Check the edges

Edges are where compositing quality shows. Look at where the product meets the background, where depth-of-field blur transitions from sharp to soft, and where colors were adjusted using identifier mattes. Clean work shows no halos, no hard cut lines inside soft regions, and no color spill from an adjusted object onto its neighbors.

Check the consistency across the series

Place all images in a series side by side. The product color, contrast and highlight character should match across them unless a difference was intended. Drift is a sign of an inconsistent color pipeline or per-image grading without a shared template.

Open the composite

If you received composite files, ask someone with the relevant software, or an independent artist, to open one. They should be able to follow the flow from passes to final, identify where each major adjustment happens, and make a simple change without breaking anything. Good composites are organized into clear stages: pass reconstruction, per-object adjustments, depth effects, global grade, output transform. Node or layer names should describe what they do. Paths to the EXRs should be relative or clearly documented, so the files open on another machine.

Test a small change

The most revealing check is to request one small change after delivery, such as a slight hue shift on a single object, and see how it is handled and how long it takes. If the pipeline is sound, this is quick and does not involve rendering. If it triggers a re-render, you have learned something important before the next project.

Depth effects, identifiers and other composite-stage decisions

Several creative decisions can be made either at render time or in the composite. Knowing where each belongs helps you judge a supplier's approach and avoid paying for render time you did not need.

Depth of field and atmosphere

Adding depth effects in the composite rather than at render time is standard good practice for most commercial stills and much animation. A depth pass records the distance from camera for each pixel, and the compositor uses it to drive a defocus or to blend in haze, fog or aerial perspective. The advantage is that focus distance, blur amount and atmosphere density become adjustable after the fact, which is exactly the kind of decision stakeholders like to revisit. Compositing applications such as Blackmagic Design's Fusion include depth-driven defocus and fog tools built for this.

There are limits. Composite defocus can produce artifacts where a sharp foreground edge sits against a blurred background, because the depth pass has only one value per pixel and cannot represent both surfaces at a partially covered edge. For extreme shallow focus with complex overlapping foreground detail, a render-time depth of field, or deep compositing data, can be the better choice. A good supplier will explain which approach they chose and why for your shots.

Object and material identifiers

Identifier passes, sometimes delivered as a set of colored mattes and sometimes as coverage-aware ID channels such as the widely used Cryptomatte approach, allow selective adjustment of individual objects or materials. They are what make the navy cap in the worked example a composite change instead of a re-render. Ask the supplier to agree the list of selectable items with you: typically the product body, each distinct material or finish, labels, props and the background. It is far cheaper to include an identifier at render time than to discover later that it was not rendered.

Light groups

Rendering the contribution of each light or group of lights to its own channel allows relighting in the composite. For product work this can be valuable, because the balance between key, fill and rim light is often debated during review. It adds render time and file size, so ask for it where lighting balance is likely to be a discussion, not by default.

When projection or background work is involved

Some projects combine rendered objects with painted or photographed backgrounds projected onto simple geometry, a technique common in architectural and environment work. If your project includes that, the composite also has to reconcile the projected background with the rendered foreground, which adds edge and lighting-match work. Our guide to camera projection mapping explains when that approach saves money and when it does not.

When a compositing project needs outside help

Some teams render in house and composite in house perfectly well. Others reach a point where the pipeline stops serving them. The situations that most often justify bringing in a specialist are clear enough to recognize.

Changes keep requiring full re-renders

If every round of feedback on your own renders means going back to the renderer, the issue is usually the pass setup and composite structure, not the artist's skill. A specialist can define a pass list and a composite template that absorbs most future changes, which often pays for itself within a single project with several review rounds.

Passes were not rendered and are now needed

This is common when a project that was expected to be a single approval turns into a campaign with variants. Sometimes the right answer is to re-render once with a proper pass set; sometimes it is careful matte work on existing beauty renders. An experienced compositor can tell you which is cheaper for your specific images, and the answer depends on how many future changes are expected.

Renders must be integrated with live-action plates

Matching CG to photography or video is its own discipline: camera tracking or matching, lighting and reflection capture, shadow catching, grain and lens matching, and color pipeline alignment with the camera's footage. If your in-house team has been compositing CG onto clean backgrounds, plate integration is the point where outside help most often makes the difference between an image that looks inserted and one that looks photographed.

Specialized output requirements

Packaging mockups, room scenes and turntables each carry their own compositing conventions. Packaging work often depends on precise label color and clean cutouts, which our guide to 3D packaging visualization covers. If you are looking for a team to handle the full chain from scene build to composite, our 3D visualization and rendering service is set up to deliver pass-based EXRs and organized composites as standard.

Writing a compositing brief that protects your budget

Pulling the handbook together, a compositing brief does not need to be long, but it should settle the questions that most often cause disputes. The strongest briefs are specific about outcomes and deliverables and leave the technical implementation to the supplier, while asking them to state it in writing.

1. Describe the images and their use. Number of images, resolution, destinations (web, social, print, broadcast), and whether future variants are likely.
2. List the changes you anticipate. Colorways, label updates, lighting mood, background treatments. Ask the supplier to map each to a pass.
3. Specify file deliverables. Final images in the stated output color space and format; multi-channel OpenEXR working files; organized composite files; a short note on how the composite is structured.
4. Specify the color pipeline. Ask the supplier to state their working space, review view transform and output transform.
5. Define revisions. Number of included rounds, how feedback will be consolidated, and what counts as a change of scope.
6. Include a test stage. One hero image taken to near-final quality, including a deliberate composite change, before the series is produced.
7. Agree archive terms. How long passes and scenes are kept, and how retrieval for future variants will work.

A brief that covers these points will draw clearer proposals, make quotes easier to compare with the scorecard above, and give you grounds to push back if a supplier later tries to charge for re-renders that the agreed passes should have made unnecessary.

Where this comes from

• Academy Software Foundation — OpenEXR and OpenColorIO
• Blackmagic Design — Fusion compositing documentation
• Pixar RenderMan — Render pass documentation

The figures and practices above come from the sources listed.

Working on something like this?

We take on 3D Design & Development work for teams who want it done once, properly. Tell us what you are building and we will tell you honestly whether we are the right studio for it. Start a project.

Where to go next

• A Practical Guide to 3D Asset Libraries and Reuse
• A Practical Guide to WebGPU and the Future of Browser 3D