How a Photo Composite Is Actually Made

A photo composite looks like a simple operation, which is why it is so often done badly. A building is placed into a photograph, the edges are tidied, and the result is presented. The steps that make it correct are invisible in the output, and skipping them produces something that looks plausible to the person who made it and wrong to anybody who knows the site.

This guide describes the actual process, in order, and what each stage determines. It is written for a client rather than a technician, because the decisions that matter most are commissioning decisions rather than technical ones.

Stage one: deciding the viewpoints

This happens before any photograph is taken and it is the decision with the longest consequences, because everything afterwards is built on it and a bad viewpoint cannot be corrected in production.

The recurring positions worth considering are the approach along the street where the proposal first becomes visible, the point directly opposite where the height relationship to neighbours is most apparent, any distant vantage from which the project will read against a skyline, and any position that a review body, a neighbour group or a public space makes important regardless of whether it flatters.

A marketing package will legitimately weight toward the flattering ones. A package intended to survive scrutiny has to include the ones that matter, and choosing deliberately between those two purposes at this stage is far cheaper than discovering the mismatch after the images exist.

Stage two: the photography, which is a site visit with rules

Somebody goes to the site and takes the photographs, and the conditions of that visit determine most of what the final image can be. This is not an errand and treating it as one is the most common origin of a weak composite.

Light direction decides whether a facade reads with relief or flat. Season decides how much street planting conceals. Weather decides the atmospheric depth that everything at distance will carry. Time of day fixes the sun position that every shadow in the final image must agree with. Traffic and parked vehicles decide whether the ground floor relationship is visible at all.

Alongside the photographs, several things should be recorded: the focal length used, the height the camera was held at, the position as precisely as available, and the exact date and time. Those four values are the inputs to the next stage, and reconstructing them afterwards is possible but lossy.

Stage three: survey, when the image will be relied upon

This is the stage that separates a defensible composite from a plausible one, and it is optional in the sense that a project can choose not to have it while accepting what that means.

A surveyor visits the same positions and records where the camera stood in three dimensions, along with control points that are visible in the photographs: building corners, kerb lines, established features. Those control points are what the virtual camera will later be aligned to.

The reason this matters is that without it, the camera position is derived from the photograph itself, which produces a solution consistent with the image but not necessarily tied to reality. A composite can be internally correct and still place the building somewhere it will not be, and nothing in the image reveals the difference.

Stage four: camera matching

This is the technical core of the discipline and it is a single operation: creating a virtual camera whose position, height, direction and focal length match the real one exactly.

With survey control, the alignment is a matter of fitting the virtual camera until the modelled control points sit on the photographed ones, which is a measurable process with a checkable result.

Without it, the alignment is solved from perspective. Lines in the photograph that are parallel in reality converge on vanishing points, and those, combined with any known dimension in the frame, allow the camera parameters to be derived. It is a legitimate technique and it inherits any error in the assumptions.

Everything downstream depends on this being right. If the camera matches, the building will be the correct size, in the correct place, at the correct apparent height. If it does not, no amount of skill afterwards will make it belong.

Stage five: the model, built to the right level of detail

The building is modelled from the drawings, and how much detail is appropriate depends entirely on how far away the camera is and how large the proposal will read in the frame.

A composite from across a river needs correct massing, correct roofline, correct proportion and the material read at distance. Detailed window frames are invisible and paying for them is waste.

A composite from the pavement opposite needs resolved openings, real material behaviour, and the details a person standing there would actually see. Massing alone will read as a placeholder.

Matching the detail to the distance is a substantial cost control lever and it is frequently ignored, with everything modelled to the same standard regardless of whether it will ever be visible.

Stage six: lighting that agrees with the photograph

The photograph fixed a date, a time and a set of weather conditions, and the render has to be lit to match them rather than to look its best. This is where the strongest temptation to cheat lives, and where the most easily detected errors occur.

The sun position is calculable from the location, date and time, and it is not negotiable: every new shadow must fall in the same direction as every existing shadow in the frame. Inconsistency here is detected by viewers who could not explain what is wrong but will not believe the image.

Overcast conditions are technically easier because the lighting is diffuse and there are no sharp shadows to reconcile, which is one reason experienced practitioners often prefer a bright overcast day for composite photography even though it is less dramatic.

Stage seven: integration, where belonging is created

A correctly aligned, correctly lit render sitting on a photograph still looks pasted on, because the two images have different physical characteristics and the eye reads those before it reads geometry.

The render is clean and the photograph has grain. The render is uniformly sharp and the photograph has depth of field. The render has no atmospheric haze and the photograph carries increasing softness and colour shift with distance. The render has perfect edges and the photograph has slight chromatic behaviour at high contrast boundaries.

Integration means reconciling all of that: adding grain to match, softening to match the depth of field at that distance, introducing the same atmospheric shift, and letting existing foreground elements pass in front of the proposal where they should.

That last one matters more than it sounds. A tree, a lamp post or a passing van in front of the building is what convinces a viewer the building is in the scene rather than on it.

The occlusion problem, and why it takes the time it does

Everything in the photograph that stands between the camera and the proposal has to be brought back in front of it, and that is manual work that scales with how cluttered the scene is.

Street trees are the hardest, because foliage has thousands of edges and each has to separate cleanly. Railings, wires, signage and vehicles all present the same problem at smaller scale.

This is frequently the single largest labour item in a composite and it is entirely invisible in the result, which makes it the item clients are most surprised to be paying for.

It is also the reason viewpoint selection has cost consequences: a clear view across an open space is dramatically cheaper to produce than an equally good view through a line of mature trees.

Where the drawings and the photograph disagree

A recurring surprise in composite work is that the site in the photograph does not match the site in the drawing set, and reconciling them is a stage nobody schedules.

Kerb lines have been moved since the survey. A neighbouring building has an addition that is not on the base plan. Street trees have been planted or removed. Ground levels at the boundary are not what the topographic survey recorded three years ago. A utility cabinet has appeared exactly where the entrance is drawn.

Every one of those has to be resolved before the render can be aligned, because the model is being fitted to a photograph of reality rather than to the drawing. Where they conflict, the photograph is correct about what exists and the drawing is correct about what is proposed, and somebody has to decide which parts of each to trust.

This is one of the quiet benefits of the format. A composite frequently discovers a discrepancy between a project assumptions and its actual site, and discovering it during image production is far cheaper than discovering it during construction.

The role of the person who knows the street

The most valuable review a composite can get is from somebody who walks past the site regularly, and it is a review almost no project arranges.

Technical checks catch geometric errors. What they do not catch is the set of things a local person notices instantly: that the light never falls that way in the afternoon, that the pavement is never that empty, that the tree on the corner is larger than that, that the building next door is a different colour than it looks in the render.

None of those are failures of process and all of them undermine credibility with the audience that matters most, which in a contested proposal is precisely the people who know the place.

Showing a draft to one such person, before the image is published, costs nothing and catches the category of error that no amount of technical rigour addresses.

Stage eight: the record

A composite produced properly leaves behind a small amount of documentation that costs almost nothing and determines whether the image can be defended later.

The camera position, the height, the direction, the focal length, the date and time, whether the position was surveyed and by whom, and the accuracy level the output corresponds to.

Kept with the image, that record answers any later question about whether the representation was fair. Discarded, the image becomes something that can only be re argued rather than checked.

What clients should review, and when

Review the viewpoints before photography, because that is the only moment they can be changed without cost.

Review the camera match before the building is fully modelled, using a simple massing block, because if the match is wrong everything modelled afterwards is wasted effort.

Review the lighting agreement before final rendering, by checking shadow direction against the photograph rather than by judging whether the image looks appealing.

Review the integration last, and review it at full size rather than on a phone, because the characteristic failures are visible at scale and invisible at thumbnail size.

The shortcuts and what each one costs

Using an existing photograph rather than commissioning one saves a site visit and inherits whatever position, light and season that photograph happened to have.

Skipping the survey saves a survey and produces an image that cannot be defended if challenged, which is fine if nobody will challenge it.

Modelling everything to the same detail regardless of distance costs money and buys nothing.

Skipping the integration stage saves hours and produces something that reads as a render on a photograph, which forfeits the credibility that was the entire reason for choosing the format.

Why this format rewards restraint

There is a temptation, once the technical work is done, to improve the photograph: brighten the sky, remove the parked cars, clean the pavement, warm the light.

Every one of those moves the image away from being a record and toward being an illustration, and the value of a composite comes entirely from the half that is a record.

The most persuasive composites are frequently the plainest ones, on an ordinary day, with the traffic and the litter and the unflattering building next door still present, because that is what makes the new element believable.

A composite that has been improved until the street looks better than it does is competing with the viewer own memory of the place, and the viewer wins that comparison every time.

Frequently asked questions

What is the most important stage in making a composite?

The camera match. If the virtual camera position, height, direction and focal length match the real ones, the building is the correct size in the correct place. If they do not, no amount of skill in later stages makes the image belong.

Why is the photography treated as a site visit rather than an errand?

Because light direction, season, weather, time of day and traffic all determine what the final image can be, and none of them can be corrected afterwards. Focal length, camera height, position and exact time should be recorded at the same visit.

What takes the most labour in a composite?

Usually occlusion: bringing everything that stands between the camera and the proposal back in front of it. Street trees are hardest because foliage has thousands of edges. It is invisible in the result, which is why clients are surprised to pay for it.

How much detail should the model have?

As much as the distance justifies. A composite from across a river needs correct massing and material read, while detailed window frames are invisible. A composite from the opposite pavement needs resolved openings and real material behaviour, where massing alone reads as a placeholder.

Should the photograph be improved?

No. Brightening the sky, removing parked cars and cleaning the street move the image from record toward illustration, and the credibility of a composite comes entirely from the half that is a record. Plain, ordinary conditions are usually more persuasive.