Build a Camera-Move Test with Separate Position and Look-At Controls
Build a Camera-Move Test with Separate Position and Look-At Controls
A treatment line like "the camera drifts past the doorway and settles on the bottle" contains two different jobs. The camera goes somewhere. The camera also looks at something. On the day, one operator does both at once, and the frame is whatever comes out of that. Change the path and you've changed what the viewer was watching; change what the viewer was watching and you've effectively redrawn the path.
In a study you don't have to accept that coupling. Build the move so position is one set of keys and aim is another, and the note "hold on the door a beat longer" stops being a note about the whole move.
That's the method, and it's small. The rest of this is how to set it up, what to look at while it plays, and what a study like this cannot tell you.
Make only the space needed to judge the move
Before you open a comp, write three sentences. What starts hidden. What interrupts the view. What becomes visible. If you can't write them, the move isn't ready to be tested — you'd be building geometry to answer a question you haven't asked yet.
Those sentences also tell you how little to build. A reveal through a doorway needs a doorway, the thing being revealed, and whatever crosses the lens. Nothing else earns its place, because anything else is a claim about a room you haven't shot in. For the walkthrough below, that's a doorway plane, a stand-in for the product, and one foreground card near the start of the travel.
Here is a placeholder stage for a hypothetical treatment. These numbers are invented scaffolding; nothing here has been rendered, and swapping in your own dimensions is part of the exercise.
- Doorway opening: roughly 900 mm wide, 2100 mm tall, standing as a simple plane.
- Product stand-in: a flat card about 100 × 250 mm, placed where the bottle would sit, a little beyond the doorway and off to one side.
- Foreground card: a sheet roughly 400 mm in front of the lens at frame one, off to the left, clearing the frame within the first second of the move.
- Timeline: five seconds at 24 fps, so 120 frames. Travel across roughly the first four seconds, hold for the last one.
Then choose your camera setting and write it down. Frame size, frame rate, which type of camera you made, the focal length you picked, and where the camera starts and ends. Record the version of the application you built it in, too. You will reopen this in three weeks with no memory of what you were testing, and the study should survive that.
Frame the beginning and the end before you animate anything. Two stills, no movement. If the last frame doesn't hold — if the product is small, off-center, or half behind the jamb — no amount of smooth travel repairs it. Movement can only carry you to a frame that was worth arriving at.
One caution that matters more than it sounds: a plausible-looking room and a familiar lens preset do not verify a location or a lens. Adobe's documentation notes that the camera in a comp is a model, an approximation of optical behavior rather than a measurement of it. What you're testing is the relationship between a few objects and a moving eye, not the room, and not the lens the job will actually use.
Animate position while the target stays still
Adobe's page on cameras and points of interest distinguishes a one-node camera from a two-node one. A one-node camera carries its own rotation values, which you author directly. A two-node camera has a point of interest, and the camera orients toward it, which means position and aim are separate animatable controls. Whichever camera is active is the one that supplies the output.
That distinction is the whole tool. Make a two-node camera, put the point of interest on the product stand-in, and key only the position.
Now look through the camera. The point of interest is what the frame centers on, and the camera's rotation is recalculated from its position on every frame so that stays true. This produces something worth pausing on: a fixed target is not a fixed frame. As the body slides forward and to the left, the aim swings to keep the stand-in centered and the doorway rotates behind it. The stand-in also grows, because you're closing on it. The travel is doing real work on the image even though the target never moved.
That's what the first pass is for. It isolates translation. Everything that moves in this version moved because the camera moved, so you can finally see what your path actually does.
It also tends to expose a problem. In this hypothetical, the bottle begins hidden behind the jamb — that's the reveal. A fixed target means the frame is centered on something the viewer cannot see yet, for the first part of the move. So the pass shows you two things at once: the shape of the travel, and the reason you'll want a second version where the aim isn't nailed down.
Then watch it from outside. Turn off the camera view and look down at the stage from above, or from the side. Check whether the path clips through the door jamb. Check how close the camera really comes to the foreground card at the moment it crosses. Find the frames where the rendering looks best and see where the camera is standing when they happen. A flattering frame can sit on top of a path that would put a real lens through a wall, or through a person.
And resist the reflex that shows up here: when a frame looks wrong, you will want to add a key on the point of interest. Do that a few times and you've rebuilt a one-node camera with extra steps, except now the aim is carrying the framing and the travel has stopped being legible as travel. If a frame is wrong, ask whether the stand-in is in the wrong place or the path is. The aim is rarely the honest answer.
Move attention only when the story needs it
Duplicate the comp and leave the first version alone. You're going to need it as the control.
In the second version, start the point of interest on the doorway detail — the jamb edge that's doing the hiding — and key a transfer to the product at the moment of discovery.
Looking through this camera, the jamb stays centered while the foreground card clears and the bottle arrives along the edge of the frame. The camera appears to be attending to the architecture while the product quietly enters. Then the aim swings and the bottle comes to center, and that arrival is the punctuation the move was built for. Notice that the bottle is legible before it is the subject. That's the arrangement you're testing.
The transfer window is narrower than it first appears, because the path constrains it. In this study the camera passes through the plane of the doorway, which means the jamb is only in front of the lens until the camera crosses it. So the aim has to leave the jamb after the product has become readable and before the jamb goes behind. Position and aim are separate controls; they are not separate facts in the finished shot. The rig is independent. The reading is not. Most of the confusion about this technique comes from forgetting the second half of that.
Then compare. Both versions have identical position keys. Only the aim differs. That's what you bought by separating the controls, and it's the comparison that justifies the extra setup — same travel, two different films.
Watch for a specific set of failures while you compare:
- A transfer that happens too early. If the aim moves to the bottle near the top of the move, the bottle is centered before it is the subject, the doorway stops withholding anything, and you have essentially rebuilt the fixed-target version with a swing bolted on. The travel goes decorative.
- A swing so fast it reads as a whip rather than a decision.
- An aim that passes close to the vertical axis through the point of interest, where the camera's sense of which way is up gets unstable and the horizon rolls or snaps.
- A reveal that resolves so quickly the viewer never registers the product arriving. If you have to point at it afterward, it didn't happen on screen.
Two limits are worth naming before you fall in love with the setup. If you need more than two or three aim positions to make the move work, the idea is probably two shots. And this rig holds an aim, not a frame: if the treatment needs the bottle to hold the same screen size while the distance to it changes — closing on it, or pulling away — a two-node camera won't deliver that, because size follows distance and aim is all the rig controls. Holding a size across a change of distance takes a distance move keyed against a focal-length change. That's a different setup, or a different idea.
Show the intended view and the assumptions behind it
Export the passage that plays through the active camera. Which camera is active matters, per Adobe's page, because the active camera is the one supplying output — a comp with several cameras and an accidental active one is a comp that won't reproduce.
Keep a second, separate record of the path: a side or overhead view with the camera's route visible, captured or recorded, showing the foreground crossing and the goal change. The clip is what the shot looks like. The path view is why.
Compare the parts that decide the move, not the opening frame. The opening is usually the most flattering frame in any study and the least informative. The foreground crossing, the moment the product becomes readable, the aim transfer, the final hold: those are the four places where a camera move succeeds or falls apart.
Hand off the editable comp with its assumptions attached. The rough geometry and its approximate dimensions. The comp size and frame rate. The application version you used. The camera type, the focal length, the start and end positions. Even if you're the only person who will ever open the file, write them down. You are about to become a different person, and that person will not remember why the camera was where it was.
Label it honestly. This is an authored viewing study — an argument about framing over time, made from flat cards and a virtual lens. It is not a solved camera, not a measured survey, and not evidence that a real rig can make the move. The stand-ins have no volume, the lens is an approximation, and nothing in the comp knows what a dolly weighs. If a physical claim survives your study — that this can be done in one move, that it fits the time available, that the space allows it — that claim needs a camera specialist and a real space to answer, not a keyboard.
End with the clip and one sentence. Not a logline; a sentence naming the attention change the clip is testing. Something along the lines of: this version holds on the door jamb while the card clears, then transfers to the bottle; the earlier version held its aim on the bottle from the first frame, so the bottle was centered before it was the subject.
That sentence is what makes the study usable in a room. Someone can disagree with an aim decision. They can't disagree with a file.
Frequently asked questions
What should be settled before building the camera-move comp?
Write three sentences first: what starts hidden, what interrupts the view, and what becomes visible. If those cannot be written, the move is not ready to be tested. Those sentences also set how little geometry to build; for the walkthrough, that is a doorway plane, a product stand-in, and one foreground card near the start of travel. The dimensions in the example are invented scaffolding, and nothing described there has been rendered.
Why make a two-node camera and key only position on the first pass?
A two-node camera has a point of interest, so position and aim are separate animatable controls. With the point of interest fixed on the stand-in and only position keyed, the pass isolates translation: everything that moves is due to the camera moving. It also exposes a likely problem—if the product begins hidden behind the jamb, a fixed target centers the frame on something the viewer cannot yet see for the first part of the move.
What is wrong with fixing a bad frame by adding point-of-interest keys?
The method warns that repeated aim keys rebuild a one-node camera with extra steps, except now the aim carries framing and travel stops being legible as travel. If a frame is wrong, ask whether the stand-in is in the wrong place or the path is; the aim is rarely the honest answer. Related limits: more than two or three aim positions probably means the idea is two shots, and a two-node rig holds an aim, not a frame, so it cannot hold the same screen size while distance changes.
How should the two aim versions be compared and handed off?
Duplicate the comp and leave the first version as the control. In the second version, start the point of interest on the doorway detail and key a transfer to the product at discovery; both versions then have identical position keys, with only aim differing. Export the passage through the active camera, keep a second path view from the side or overhead, and hand off the editable comp with its assumptions: rough geometry and dimensions, comp size and frame rate, application version, camera type, focal length, and start and end positions. Compare the foreground crossing, the moment the product becomes readable, the aim transfer, and the final hold—not the opening frame. Label it as an authored viewing study, not a solved camera or evidence a real rig can make the move.