Build an Exploded-View Test From a Supplied Product Assembly

An exploded-view test is a short animation of parts you already have: they come apart far enough to show how they relate, then go back exactly where they started. For a treatment, that sequence does one job — it makes a proposed relationship visible in a form the team can scrub, re-time and argue with.

The order of work matters more than the tool. Record where every part sits before anything moves. Decide the one relationship the shot has to reveal. Separate the relevant components along directions you chose on purpose, at times a viewer can follow. Then put them back and compare the result against the recording, not against how the final frame feels.

Two things the study can never do, however good the animation gets. It cannot show the inside of a part you were not given. And a separation path that reads well on screen is not evidence that the product comes apart that way.

What the study can and cannot settle

An exploded view moves existing parts. That is the whole mechanism. It takes components that are already separate in the file, gives them travel, and lets a viewer see the arrangement that the assembled state hides.

It is not a cutaway. A cutaway implies you know what is behind the surface. It is not a teardown, because a teardown is a claim about sequence, force and access on a real line. And it is not a morph: if the treatment wants product A to become product B, that is a different construction and it should not be sold as an exploded view of either one.

What you can honestly deliver is narrower and more useful: these are the parts we were given, this is where they sit when assembled, and this is the relationship the shot will make legible. Everything after that is craft.

Read the supplied assembly before you move anything

Open the model and look at the tree before you look at the viewport. Three questions decide whether the study is even possible.

Are the parts actually separate? If the shell, insert and lid arrive as one fused body, you do not have an exploded view waiting to happen. You have a modeling decision. Splitting geometry by eye and calling the result the product's construction is inventing anatomy, and the treatment should not carry it as if it were supplied.

What is a component and what is a body? The thing you animate needs to be the node that owns the geometry, not a fragment of it. Read the names in the hierarchy. If the names are generic — Body 14, Body 15 — that is worth resolving with whoever supplied the model before you build motion on top of a guess.

Which relationship does the shot have to reveal? Pick one. In the fixture below, the answer is containment: the insert sits inside the shell, entirely below the rim, and nothing in the assembled state shows that. One relationship, stated in a sentence, is what the timing, the directions and the camera all get built to serve. "Show how it's made" is not a relationship. It is a wish.

Record home positions as numbers

Before you move a single part, write down the home state in a form you can compare against later. A screenshot is useful. A screenshot is not a transform.

Get three things per component: its name, its position and rotation relative to the assembly origin, and which node owns it. If the file already has a saved home state, confirm what that state actually is rather than assuming it matches the assembled view you are looking at.

Autodesk's Fusion help documentation describes a Transform Components operation, a manual explode, and a Restore Home action that returns components to the location recorded in the Design workspace (help page). The provider's own tutorial describes the animation workspace as keeping component movement, camera changes and sequence timing in separate lanes (tutorial — the article, not its linked video, which I have not watched).

Those are documentation claims, and the one that matters most here is also the one you should not take on trust. "Restore Home returns the part exactly" is precisely the claim your fixture exists to test. Run it on your own model, in a named version, and look at the result from two angles before you believe it.

One more thing worth doing at this stage: separate component motion from camera motion in your head, and then in the timeline. A camera move that happens while parts are flying makes it much harder to tell whether the parts moved correctly. Hold the camera for the separation pass. If the treatment needs a move, build it as a second pass over the same animation.

Stage the separation — a fixture to reason with

The fixture below is invented. It is three simple parts with round numbers, written down here so the checks are checkable. Nothing in it was cut, modeled, animated or rendered; it is a paper construction, and it does not describe any real product's mechanism.

Each component's origin is the center of its own bottom face, and the assembly origin is the center of the shell's base. Assembled, the shell occupies z 0–30, the insert z 4–24, and the lid z 30–34, with the boss at z 34–35.5. The boss is centered at x = 0, y = +14 — toward the back edge of the lid, since the lid runs y = −20 to +20.

The insert sits 1 mm clear of the cavity walls and its top face is 6 mm below the shell's rim. That is why the relationship is worth a shot: from a low front angle, the shell's front wall hides the insert completely.

Two ways to separate the same three parts

The simplest separation moves everything at once. Lid up 40 mm, insert up 36 mm, both over the same 1.2 seconds, shell held still as the ground reference. The numbers work: the lid ends at z 70–74, the insert at z 40–60, and the gap between them is 10 mm at rest. Even mid-move the two never intersect, and the insert clears the shell rim about 72 percent of the way through.

Nothing is wrong with that version. It is just harder to read, because the viewer has to track two moving parts and infer two relationships at the same moment.

The staged version gives each relationship its own beat. A six-second cut:

• 0.00–0.90 — lid up 40 mm
• 0.90–1.40 — hold
• 1.40–2.60 — insert up 36 mm
• 2.60–3.20 — hold, fully separated
• 3.20–4.40 — insert down 36 mm
• 4.40–4.80 — hold
• 4.80–5.70 — lid down 40 mm
• 5.70–6.00 — hold at home

The first hold is doing real work. It lets the viewer register that the lid is gone and the top of the shell is now open before the insert starts rising. Without it, the two moves blur into one gesture and the sequence explains less than the all-at-once version while taking longer.

Three judgments run through this part of the build. Direction first: the insert leaves straight up because the cavity is a straight-walled open box — but that is a property of this fixture, and it says nothing about a real part with an undercut, a snap feature or a lip that has to rotate out. Second, distance: 10 mm of clearance is enough to read the split and close enough to see that the parts belong together. Pushing everything to the far corners of the frame reads as scatter, not as structure. Third, pivots: every motion here is a translation, but the moment you rotate something, the pivot decides what the rotation does to the part's position. In this fixture the lid's pivot is the center of its own bottom face, which is why a half-turn swaps the boss end-for-end without moving the plate. Move the pivot and the same rotation also slides the part.

Label the paths. A direction chosen because it reveals a component is an illustrative path. It is not a disassembly route, and the treatment should not imply that it is.

The return is the test

Bring the parts back and compare the final state against the recording you made at the start. Not against the previous frame, which will look fine, and not against the front view alone.

Here is the failure the fixture is built to catch. On the return, the lid comes down to (0, 0, 30) correctly, but it arrives rotated 180° about its own Z axis. The plate is 60 × 40 and centered on its origin, so a half-turn maps its footprint onto itself. The height is right. The seat on the shell rim looks right. The only thing that changed is the boss, which has moved from y = +14 to y = −14 — from the back edge of the lid to the front edge.

In the low front view the temptation is to accept it. At a shallow camera height the lid's top face compresses into a sliver, the boss reads as a small bump somewhere along the top edge, and the silhouette is identical to the correct version either way. Nothing in that frame announces a problem.

A plan view settles it in one look. Drop the camera to top-down and the boss is at the wrong end of the lid, 28 mm from where it should be.

That is the whole lesson of the exercise, and it generalizes past this fixture: the angle you liked for the hero frame is often the angle that hides the mistake. Keep at least two informative views of the home state in the study — a three-quarter view for shape and a top-down or side view for placement — and check the return in both. When the transform is available as a number, compare the number too. A rotation that should read as zero and reads as 180 is a settlement no amount of rendering will make.

If the lid had been a plain plate, no view would have caught it and the number would have been the only witness. That is worth remembering the next time the part is symmetric.

Save the study with its limits attached

Hand over four things together:

• The native file, with its edit history and timeline intact, so the timing can be re-cut without rebuilding.
• The home-state recording — the positions and rotations you wrote down before anything moved.
• At least two views of the separated and returned states, from the angles you used to check.
• A short list of what the motion does not claim. Which paths are illustrative. Which parts were fused and remain unresolved. Which product details you were never given.

Label illustrative movement where the viewer will see it. A caption on the file, a note in the treatment, a differently colored marker in the timeline — the form matters less than that the annotation travels with the animation rather than living in someone's memory. The same goes for appearance: setting a material so a part reads clearly against the shell is a presentation choice. It does not tell anyone anything about the part's engineering properties, and it should not be read as if it did.

What you hand over is then a small, honest, editable thing: the parts we have, moving apart far enough to show one relationship, and coming back to exactly where they were. It survives the next round of notes because the changes are cheap. And it survives scrutiny because the claim stopped where the supplied model stopped.

Frequently asked questions

What is an exploded-view test, and what is it not?

It is a short animation of parts already in a supplied assembly: they come apart far enough to show how they relate, then go back exactly where they started. It is not a cutaway, because a cutaway implies knowledge of what is behind the surface. It is not a teardown, because a teardown claims sequence, force, and access on a real line. It is not a morph between different products.

What should be checked in the supplied assembly before animating?

Look at the tree before the viewport. Check whether the parts are actually separate, since fused geometry is a modeling decision rather than an exploded view waiting to happen. Identify what is a component and what is a body, and resolve generic names if needed. Then pick one relationship the shot has to reveal. Also record home positions as numbers per component: name, position and rotation relative to the assembly origin, and which node owns it.

What is the difference between all-at-once and staged separation?

All-at-once moves everything at the same time. The numbers may work, but it is harder to read because the viewer tracks two moving parts and infers two relationships at once. A staged version gives each relationship its own beat, with holds before and after each move so the viewer can register that, for example, the lid is gone and the shell's top is open before the insert starts rising.

Why is the return the real test of an exploded-view study?

The final state should be compared against the recorded home state, not against the previous frame or the front view alone. In the fixture, the lid returns at the correct height but rotated 180°, moving its boss from the back edge to the front. A low front view hides this because the top face compresses and the boss reads as a small bump, but a top-down view shows the boss at the wrong end. Keep at least two informative views of the home state and compare the transform numbers too when available.

What limits should travel with the saved study?

Save the native file with edit history and timeline, the home-state recording of positions and rotations before anything moved, at least two views of the separated and returned states, and a short list of what the motion does not claim. Those limits include which separation paths are illustrative rather than disassembly routes, which parts were fused and remain unresolved, which product details were never supplied, and that appearance or material settings are presentation choices rather than engineering evidence.

Further reading

• Pitch an Exploded-View Commercial Without Inventing What Is Inside the Product — Read this when the build becomes a pitch line: it shows how movement order and spacing read as claims, and which parts you may honestly separate.
• Fold a Supplied Packaging Dieline Into a Product Mockup for a Treatment — Read this for the same discipline on a supplied dieline: inspect a list of views rather than a hero view, and say what the folded result cannot certify.