Previs has a reputation problem. Say the word and most filmmakers picture a room of 3D artists in Maya, a game-engine pipeline, and a line item no independent production can carry. But previs is not a rendering technique. It is a set of decisions: where the camera goes, what lens it wears, how the actors move, what coverage you need, and how long the scene should run. Any workflow that lets you make those decisions before the shoot day is doing previs work, and you can run the entire process with boards, a timed animatic, and AI-generated motion. No 3D software, no specialist hires. Here is the workflow, step by step.
What previs is actually for
Strip away the tooling and previs exists to answer questions while answers are still cheap. Does the blocking work in this location? Does the scene need seven setups or four? Does the oner hold interest for its full length, or does it sag in the middle? On set, each of those questions costs crew time to answer. In prep, they cost nothing but your attention. Traditional 3D previs answers them by simulating the scene in space. The workflow below answers the same questions with less machinery, and for most scenes the answers are just as reliable.
Step 1: Break the scene down on the page
Start with the script, not the shot ideas. Read the scene and mark its beats: every shift in intention, every entrance and exit, every piece of business that must be seen for the story to track. Note the emotional turn, the single moment the scene exists to deliver, because your most deliberate camera decision should land there. List what the frame must contain at each beat: the letter on the table, the gun in the waistband, the reaction the audience needs.
This pass takes twenty minutes and shapes everything downstream. A scene broken into six beats tells you the minimum structure your coverage has to serve. Skipping this step is how filmmakers end up with beautiful boards of the wrong moments.
Step 2: Draft a working shot list
Turn the beat breakdown into a coverage plan: for each beat, decide the shot size, angle, movement, and a first-guess lens. This draft shot list is a hypothesis, not a contract. You will revise it against the boards in step 3, and the final version your AD schedules from should always come off the finished boards.
Make the lens calls now, even roughly, because focal length is a storytelling decision, not a technical afterthought. A 24mm close-up and an 85mm close-up are the same shot size and completely different sentences: the wide lens exaggerates space and puts the audience inside the character's bubble, the long lens compresses the background and observes from a distance. Deciding at this stage whether the scene lives on wide glass or long glass gives every board a point of view. If you want a structure to work inside, start from a shot list template rather than a blank page.
Step 3: Board every setup, and draw the floor
Now make the shot list visible. One panel per setup, framed the way the lens would actually see it. This is where AI generation replaces weeks of drawing or the 3D modeling you are avoiding. In mStudio, you can import your screenplay as a PDF and generate a first-pass board for the whole scene, then regenerate individual frames until the composition matches your intent, with consistent characters held from setup to setup so the boards read as one film rather than a mood collage. If you have location scouts or reference photography, upload those as frames alongside the generated ones. An AI storyboard generator earning its keep should let you describe framing in lens terms, a low-angle 35mm two-shot, a long-lens single against a compressed background, and return something frame-accurate enough to make the coverage decision.
While you board, draw the floor. A simple overhead blocking diagram per scene, actor positions, movement paths, and camera positions marked, catches the errors that sink shoot days. Draw the axis of action between your two principal characters and keep your camera positions on one side of it: cross the 180-degree line mid-scene and your characters swap screen direction, and the coverage stops cutting. If the blocking demands a line cross, plan the bridge, a neutral shot on the axis or a visible camera move across it, and board that bridge explicitly. Check eyelines panel to panel while you are at it. These are exactly the spatial mistakes 3D previs was invented to catch, and a diagram plus disciplined boards catches them for free.
Step 4: Cut a timed animatic with scratch audio
Static boards cannot tell you how long the scene is, so put them in time. Sequence the panels on a timeline and give each shot its intended duration. Then add scratch audio, because pacing without sound is guesswork: a voiceover reading the dialogue, a temp music cue for tone, and the two or three sound effects that anchor the scene's hits. In mStudio the animatic stage lives in the same workspace as the boards, with generated voiceover, music, and sound effects you can trim, fade, and mix per track, so timing the scene does not mean exporting to an editing app.
Watch the result at full length, in real time, without touching the keyboard. This first viewing is the closest thing to dailies you will get before the shoot, and it will surprise you.
Step 5: Iterate the coverage like an editor
Now interrogate the cut. Where did your attention drop? That shot is too long or should not exist. Where did you need a detail the boards never showed, the hand on the doorknob, the reaction from across the room? That is a missing insert, and finding it now costs one generated frame instead of a pickup day. Which setups never earned a cut? Kill them and give their minutes to the shots that matter. Adjust durations, re-board what needs re-boarding, and run it again. Two or three passes is normal, and version history on each frame means you can walk back a change that seemed smarter than it was.
This loop, watch, diagnose, revise, is the actual substance of previs. The 3D pipeline was only ever a way to feed it.
Motion for the shots a still cannot answer
Some questions survive the animatic. Does the crane move from the floor to the balcony hold together? Does the fight choreography read at this camera height? A static frame cannot say, and this is where productions used to either hire a previs artist or gamble. Image-to-video generation closes the gap: take the board frame and generate the move, then judge the result the way you would judge a blocking rehearsal. It is not techvis, and it will not hand your key grip measured crane data or camera-tracking exports. But for the creative half of the question, whether the shot works, whether it cuts against its neighbors, whether it justifies the setup time, generated motion is frame-accurate enough to decide with confidence, which is the decision previs exists to enable.
The whole pipeline, one afternoon
Break down the scene, draft the shot list, board every setup with a floor diagram, cut a timed animatic with scratch sound, then iterate until the scene earns every one of its shots. A workflow that once required weeks of 3D artistry now fits between a morning coffee and a production meeting, and it produces something a Maya scene never did: a watchable cut of your scene that the entire crew understands on sight.
mStudio runs this whole pipeline in one place, screenplay import to boards to timed animatic to generated motion, with an mp4 export for the production meeting. If you have a scene that needs answers before the shoot day, start it in our previs software and see how many decisions you can lock this week.