What is previsualization?
Previsualization — previs — is working a shot out in rough 3D before committing to it. Grey boxes standing in for actors and sets, a real camera moving through them, cut together at the intended length. It answers "does this sequence work" while the answer is still cheap.
What it is, precisely
Previs is deliberately ugly. Nothing is textured, lit or finished, because appearance is not what is being tested. What is being tested is staging — where things are relative to each other — coverage, meaning which angles tell the story, and timing: how long each shot holds and where the cuts land.
That restraint is the point. A previs that looks finished invites notes about the wrong things. A previs made of blocks gets notes about whether the sequence reads.
Where it came from
Storyboards did this job for most of film history, and still do. Previs adds the two things a drawing cannot carry: real camera geometry, so a lens and a distance produce a frame you can trust, and real duration, so a cut has a length rather than a caption saying "beat".
It became standard on effects-heavy productions because those shots are expensive to get wrong. Deciding a crane move does not work costs an afternoon in previs and a day of crew time on the day.
Why generative video brought it back
Text-to-video inverted the economics. A shot now costs a credit and a minute rather than a crew and a day — so nobody plans, they simply regenerate. That works right up to the point where the thing you cannot get is a specific camera move, and no amount of regenerating produces it, because the instruction was never carried in the first place.
Prompts describe subjects well and camera work badly: the reason is structural, and it does not improve by writing a longer prompt. A previs clip sidesteps it. The motion is shown rather than described, and a video model given a reference has nothing left to interpret about where the camera goes.
So previs returns for a reason that would have sounded strange a few years ago: not to save money on a shoot, but to be the one artefact that can state a camera move unambiguously.
What previs is not
- Not animatics. An animatic is storyboards cut to timing. Previs has a real camera in a real space, so you can change the lens and see what happens.
- Not a finished render. If it looks finished, it took too long and it will attract the wrong notes.
- Not concept art. Concept art settles how something looks. Previs settles where it is and when you see it.
What a useful previs contains
- Objects at true scale. A person is 1.8 m and a car is 4.4 m long. If the scale is wrong, every framing derived from it is wrong too.
- A camera with a real lens. Focal length, position and aim, so the frame you see is the frame you would get.
- Motion on a shared clock. The action runs on one timeline and each shot is a window onto it, which is what lets two angles cover the same moment.
- Cuts at their real length. A sequence that reads well at four seconds a shot may not at two.
Frequently asked
What does previs mean in film?
Previsualization: planning a shot or sequence in rough 3D before shooting it, using untextured placeholder geometry and a real virtual camera to test staging, coverage and timing.
What is the difference between previs and storyboards?
Storyboards are drawn stills. Previs is moving 3D with a real camera, so it carries lens choice, camera path and shot duration — things a drawing can only annotate.
Is previs the same as an animatic?
No. An animatic is storyboard frames cut to timing. Previs is built in 3D, so the camera can be moved and the lens changed and the frame updates accordingly.
Why does AI video need previs?
Because prompts carry subject description reliably and camera direction poorly. A previs clip shows the camera path and timing directly, which removes the ambiguity a prompt cannot resolve.
Do I need previs if I am only making short clips?
Only if the camera has to do something specific. For a static or incidental shot a prompt is enough. For a named move, a sequence of moves, or timing tied to action, a reference clip is the difference between asking and getting.
Blockshot is previs built for this: describe a scene, get a blockout with real camera moves, export the clip. See how it works →