Previs artist

What they do

Previs artists build a scene in 3D before it is shot, at a fidelity deliberately far below final: grey models, blocked characters, a real camera with a real lens, and timing that runs at the length the shot will run. Input is a storyboard, a script page, a set survey, sometimes a stunt rehearsal on video. Output is a cut sequence with a frame count, delivered to the director, the cinematographer, the VFX supervisor and the first AD. Much of the value is not the picture. Techvis takes the same scene and returns numbers the floor can use: crane reach, how much bluescreen is in frame at a 32mm, the exact truck-in speed a motion-control rig has to hit, whether the set has to be built to the top of the wall. Postvis, later, cuts rough CG into plates already shot so the editor can cut a sequence that has no finished effects in it yet. Tool choice has shifted toward real-time: Unreal Engine work with game-style assets, alongside Maya, and the department often sits inside a VFX facility rather than at the studio. The unglamorous fact of the job is that previs is thrown away. Nothing in the file goes to final. Its purpose is to make an expensive decision cheaply, then be discarded once the decision is made.

What they decide

The calls that belong to this role and no other.

• How a boarded sequence is broken into camera setups, proposed to the director for approval
• The lens, height, and camera path in the previs, which the cinematographer may adopt, modify, or ignore
• What fidelity each shot needs, since anything past readable is wasted work
• The timing and frame count of each shot, which drives schedule and stage estimates
• Which technical constraints get surfaced in techvis: rig reach, set extent, greenscreen coverage, plate size
• Where a real-time engine is faster than a traditional 3D package for the current problem

Principles of the craft

• Previs is a decision tool with a shelf life: the shot exists to settle a question, then be deleted. Judge every hour against whether it changes what somebody does on the day, and stop the moment the answer is clear.
• Fidelity to the question, not to the render: if the question is whether the helicopter clears the bridge, grey geometry answers it. Detail added past readability costs schedule and, worse, invites notes on things that will not exist in the final image.
• Real lenses or nothing: previs done at a made-up field of view is a lie that reaches the floor. Match sensor format and focal length to the camera package so a 27mm in previs is a 27mm on the day.
• Techvis is where previs earns its budget: the diagram that says the crane needs another two metres of reach saves a shooting morning. Deliver numbers, not just pictures, and deliver them before the build is locked.
• Previs is a proposal, not a plan: the director and the cinematographer own the shot. Offer three stagings and reasons rather than one, and expect the good version to be the one they change.

Common mistakes

• Polishing previs into a finished-looking piece, which burns days and gets the department notes on lighting and texture nobody asked it to solve.
• Making shots that cannot be executed: impossible camera positions, moves no rig can hold, or a lens that puts a stage wall in frame.
• Treating previs as final layout. It is not built for it, and handing it downstream as if it were pushes broken cameras into the pipeline.
• Skipping frame counts and timings, so editorial cannot cut the sequence and production cannot cost it.
• The competent artist's version: falling in love with a camera move and defending it, when the point of the department is to make alternatives cheap.

How to learn it

Exercises and routes in, not a reading list.

• Previs one action beat from a screenplay you did not write, in Blender or Unreal with grey boxes and no textures, in three camera setups. Set the virtual camera to a real sensor size and only use 24mm, 40mm and 85mm. Cut it to length and note what the lens choice changed about the geography.
• Take a previs you built and write the techvis for it on one page: crane height, dolly length, how much of the frame is greenscreen at each setup, and the shortest lens that keeps the set edge out of frame. Then break your own previs by putting a stage wall where it would really be.
• Rebuild a sequence from a released film as previs, timed to the actual cut length. Compare your camera positions to the finished shots and note where you covered the geography and they did not.
• Read the fxguide previs features, then previs the same action twice: once in Maya, once in an engine. Time yourself. The gap tells you where real-time actually pays.
• Take a finished previs, strip everything past grey boxes, and check whether it still answers the question. Whatever you can delete without losing the answer should never have been made.