Make a Simple 3D Location Blockout for a Film Pitch

A location photograph answers a question you have already framed. It shows the corner, the ceiling, the light, the way the room sits in its building. What it cannot show you is whether a character at the far table will see who walks in through the door — and that is usually the question the scene is actually built on.

The tool for that question is a blockout: a plain volume model of the proposed space, built from a few boxes, plus a small set of views that all describe the same geometry. Its value is not realism. Its value is that someone can point at a view and say, "there — that's the problem," and then you can move one thing and check again.

Two habits keep a blockout useful, and most of this article follows from them. Build only the volumes your question needs. And label every dimension by how you got it.

The second habit is the one people skip. A rough model full of unlabeled numbers looks like a survey and isn't one. If a wall height came from a plan you were given, say so. If you paced it out, say that. If the partition does not exist yet and you invented it to test an idea, say that too, in the model, on the views, and in the caption of every image you hand over. The moment a reader can't tell which numbers are facts about a building and which are a proposal, the blockout stops being evidence and starts being decoration.

Start from a plan you're allowed to use, and write down one question

You need a footprint — yours, or one you have permission to work from. It can be a plan you drew, a sketch from a scout, a plan a location supplied with the understanding that you'd use it for prep. What matters is that you know its measurement status before you start drawing, because everything downstream inherits it.

Then write the question in a single sentence that a blockout can actually answer:

• Can someone standing at the far end of the room see the doorway past the partition?
• If the table moves half a meter, does the actor at the window still have a clear path to the door?
• Does the second entrance sit inside or outside the sightline from the hallway?

Those are geometric questions. "Does the room feel oppressive" is not, and a blockout will not settle it. Framing the question first is what lets you delete most of the model later.

One distinction is worth stating plainly, because it comes up in every pitch conversation: the story space and the filming location are not necessarily the same thing. A blockout describes a proposed space — what your script needs the room to do. It is not a claim that a particular building has been secured, and it is not a survey of one. If your numbers come from a real place, say which place and how you got them. If they are invented, that is fine; just don't let the invented ones borrow authority from the measured ones.

The example: one room, one doorway, one movable partition

This example is calculated from its coordinates rather than captured from a finished model, so treat its visibility answers as arithmetic you can check, not as a demonstration someone performed. Building it is the point, and it should take about twenty minutes.

A rectangular room, 6.0 m × 4.0 m in plan, 2.7 m to the ceiling. Set X from 0 at the west wall to 6.0 at the east wall, and Y from 0 at the north wall to 4.0 at the south wall. One doorway in the west wall, 1.0 m wide, spanning Y = 1.5 to 2.5. A movable partition parallel to the west wall, standing 3.0 m in from it, 2.0 m long, 2.2 m tall, 0.1 m thick, spanning Y = 0.5 to 2.5. Two observer positions at the east end: A at (5.0, 1.0) and B at (5.0, 3.5), both at an eye height of 1.6 m. Every number here is invented, and in your own model each one should carry a label:

On a real job, some of those rows would read measured — named plan, date received — and some would read estimated. Estimated is not a softer word for measured, and it should never be rounded to look precise. Pull a ceiling height from a photograph and you have guessed; write "est. 2.7 m" and the blockout stays honest about it.

Build only the volumes the question needs

SketchUp's Push/Pull tool does one thing that covers almost all of this: it extrudes a face into volume and accepts a typed distance rather than a drag (Pushing and Pulling Shapes into 3D, page updated 2026-08-21). Draw the floor rectangle at 6.0 × 4.0. Draw each wall as a narrow footprint rectangle on the ground plane and pull it up 2.7 m. Wall thickness doesn't matter for this study — pick 0.15 m and keep it identical on both passes, and let the corners overlap slightly. You are not producing a set drawing.

For the doorway, you can avoid cutting a hole entirely: build the west wall as two pieces, one running Y = 0 to 1.5 and one running Y = 2.5 to 4.0, with the opening between them. If you want the opening to read like a door rather than a full-height gap, pull a third rectangle to 0.6 m and move that group up 2.1 m to become the header. It doesn't change this test — the sightlines that decide the answers below all cross the partition below its 2.2 m top — but it makes the eye-height views easier for a non-modeler to read.

The partition is its own object. This is where grouping does real work: groups isolate geometry from other entities and allow independent editing (Grouping Geometry, page updated 2026-08-21). Make the floor, the four walls, and the partition separate groups and name them. Wall_W, Wall_N, Partition. Naming sounds fussy until the moment you slide the wrong object and stretch a wall into the room — which is exactly the failure this experiment is designed to survive.

Leave out everything else. No ceiling, so the top view reads as a plan. No furniture, no trim, no door leaf. If a piece of furniture is what your scene turns on, model that one thing and nothing around it.

Establish an overhead reference before choosing a view

Switch to a top view with parallel projection. SketchUp distinguishes standard views and projection choices, and parallel projection is the one that makes the plan read at true proportions (Viewing a Model, page updated 2026-09-17). Perspective, which you'll use in a moment for the eye-height views, makes parallel lines converge — good for showing a space, wrong for measuring one at a glance.

On this overhead view, label the doorway, the partition, and the two observer positions. Use the same letters on every view that follows. If you mark A and B on the plan, A and B mean the same two places in every image you export.

Label the dimensions with their status while you're here. A line of text on the ground plane saying "6.0 m — invented" costs nothing and travels with the model.

What the overhead view is not: a record of where you were looking. The pleasing three-quarter orbit you drifted into while modeling is a working view, not a specified one, and it will not reproduce.

Place two viewpoints deliberately

Position Camera sets the camera at a chosen point, and the Eye Height setting fixes how far above the ground plane it sits (Walking through a Model, page updated 2026-08-21). Put the camera at A — (5.0, 1.0) — set eye height to 1.6 m, and aim it at the doorway. Save that view. Then do the same at B.

Three settings need to match between the two views or the comparison is meaningless: eye height, field of view, and whether you're in perspective or parallel projection. Whatever your edition calls the framing setting — field of view or focal length — set it once for A and leave it alone for B. A wide view at A and a narrow one at B compares two different cameras, not two positions.

Both views should show the same doorway and the same partition. If one of them doesn't have the doorway in frame, you've framed a different question.

And keep the claim narrow. A model view shows what the geometry permits; it does not show that a real camera and lens can stand at that point, or that a crew can get behind it. Those are location questions, and they belong to a scout, not to a blockout.

Move the partition, not the camera

Here is the test, and here is the reason to have done all of the labeling first.

In the baseline state — partition at X = 3.0, spanning Y = 0.5 to 2.5 — observer A sees none of the doorway, and observer B sees all of it.

The reasoning is short enough to check by hand. The partition is 2.0 m in front of each observer and the doorway is 3.0 m beyond it, so every sightline crosses the partition's plane two-fifths of the way from the observer to the doorway. A's sightlines to the two edges of the opening cross at Y = 1.2 and Y = 1.6 — both inside the partition's 0.5-to-2.5 span, so the whole opening is hidden. B's cross at Y = 2.7 and Y = 3.1, both outside that span, so B sees through the gap to the south. Because the edges of the doorway bound every sightline to it, checking those two lines settles the entire opening.

Now move the partition 1.5 m south, so it spans Y = 2.0 to 4.0 and its south end meets the south wall. Change nothing else. Reopen the overhead view and both eye-height views.

The answer inverts. A's crossing points — still 1.2 and 1.6, because A has not moved — now sit north of the partition, so A sees the whole doorway. B's crossing points — still 2.7 and 3.1 — now sit inside the partition's span, so B sees nothing.

If you slide the partition only 1.0 m instead, to Y = 1.5 to 3.5, you get a partial result: A's crossing at 1.2 clears the partition, A's crossing at 1.6 doesn't, and A sees roughly the northern three-quarters of the opening. That intermediate state is worth looking at once, because it shows the thing a blockout is actually good at. The answer doesn't arrive as a yes or no. It arrives as a threshold, and you can watch the threshold move.

Two errors will make this experiment lie to you. The first is moving the wrong group: if a wall shifts instead of the partition, the comparison is no longer about the partition, and you'll have a room that doesn't match the plan you drew it from. The second is moving the camera between passes and calling it a geometry change. Moving a camera changes a picture. Moving geometry changes the space the picture is of. Either can be the right move for the pitch — but they answer different questions, and only one of them is this test.

Export views that stay accountable to the model

The export route varies by edition, and that part hasn't been verified here, so check it in your own copy before you promise anything to anyone. The requirement is simpler than the mechanism: the native model is the source, and every exported image has to be reproducible from it. Reopen the file, restore the same views, and compare before you deliver.

Each image gets a caption naming the relationship it explains — "Baseline: doorway hidden from A" — and the label status of the dimensions that matter to that relationship. Keep the model's own labels visible in the frame rather than tidy in a separate document. The images are what leaves the room; the model stays on your laptop.

Don't retouch an export into a space the source model doesn't contain. Widening a view or tidying a wall in an image editor produces a picture that no longer answers to the geometry, and the whole value of a blockout is that the two agree.

Saved view states are the obvious way to make a view set reproducible, and the documentation this piece worked from covers the camera controls rather than scene management, so treat that as yours to check.

What you have at the end

Two states of one model, a matched view set for each, and the one spatial decision they support: with the partition in its original position, the character standing where A stands cannot see who enters. If the scene needs A to see the entry, the partition has to move about 1.1 m south — enough for A's sightlines to clear its north end — and 1.5 m puts it flush with the south wall. Or A has to stand somewhere else.

That is a sentence you can say in a pitch meeting, and it is worth more than a beautiful render of a room nobody has asked a question about. When the arithmetic surprises you — when the doorway turns out to be visible from a position you assumed was blind, or hidden from one you assumed was clear — say so. A labeled uncertainty is a useful result. Decorative realism is not proof of anything.

Operations referenced above come from four SketchUp help pages: Push/Pull (updated 2026-08-21), Viewing a Model (updated 2026-09-17), Walking through a Model (updated 2026-08-21), and Grouping Geometry (updated 2026-08-21), the last of these inspected on 2026-09-18. These pages document the modeling and viewing mechanisms described; they don't guarantee that a physical camera and lens can occupy any position shown, and no model, export, or measured location was used to produce the example's numbers.

Frequently asked questions

What should a blockout include, and what should it leave out?

Build only the volumes your question needs: a plain volume model of the proposed space, made from a few boxes, plus a small set of views that describe the same geometry. Leave out the ceiling so the top view reads as a plan, and omit furniture, trim and door leaf unless one piece of furniture is what the scene turns on; then model that one thing and nothing around it. Make the floor, walls and partition separate named groups. Wall thickness does not matter for this study, so pick a value and keep it identical.

What kind of question can a blockout actually answer?

It can answer a geometric question written in one sentence, such as whether someone at the far end of the room can see the doorway past the partition, whether moving the table half a meter still leaves a clear path to the door, or whether a second entrance sits inside or outside a sightline from the hallway. 'Does the room feel oppressive' is not a geometric question, and a blockout will not settle it. The blockout describes a proposed story space; it is not a claim that a particular building has been secured and not a survey of one.

Why label every dimension by how you got it?

A rough model full of unlabeled numbers looks like a survey and is not one. If a wall height came from a plan you were given, say so; if you paced it out, say that; if a partition does not exist yet and you invented it to test an idea, say that in the model, on the views and in every caption. Estimated is not a softer word for measured, and it should not be rounded to look precise. Once a reader cannot tell which numbers are facts about a building and which are a proposal, the blockout stops being evidence.

In the article's example, what happens when the partition moves?

Baseline partition at X = 3.0 spanning Y = 0.5 to 2.5: observer A sees none of the doorway, and observer B sees all of it. Move the partition 1.5 m south, to Y = 2.0 to 4.0, and the answer inverts: A sees the whole doorway and B sees none. Slide it only 1.0 m, to Y = 1.5 to 3.5, and A sees roughly the northern three-quarters of the opening. If the scene needs A to see the entry, the partition has to move about 1.1 m south, enough for A's sightlines to clear its north end; 1.5 m puts it flush with the south wall, or A has to stand somewhere else.

What settings must match between the two eye-height views, and what are their limits?

Eye height, field of view or focal length, and whether you are in perspective or parallel projection must match between the two views. Both views should show the same doorway and the same partition. A model view shows what the geometry permits; it does not show that a real camera and lens can stand at that point, or that a crew can get behind it. Those are location questions and belong to a scout, not to a blockout.

Further reading

• Pitch a Film’s Geography, Not Just Its Locations — Read this first when the pitch still has to name the sightline, route or boundary the scene depends on, so the blockout has a relationship to model.
• Read a Set Plan and Elevation Before Describing the Commercial's Space — Use this when the plan you are modeling from arrives with a title block, so you can tell what the drawing's status lets you claim about the room.