Storyboarding

This framework was generated using the Framework Builder methodology. It demonstrates the five-layer DNA architecture applied to storyboarding short-form video: ads, explainers, social spots, and scripted scenes.

It is for anyone who has to hand a video plan to someone else and have them execute it: a solo operator boarding their own shoot, a small team briefing an editor or animator, or someone directing a generative video tool. A model asked to "storyboard this" writes panel descriptions. This produces a production document.

Layer 1: Principles Foundation

Principle 1: A storyboard is a production document, not a set of pictures. Its job is to let someone who was not in the conversation shoot or cut the thing without calling you. The test is not whether the drawings look good, it's whether a shot list, a prop list, and a runtime can be read straight off the board. Pretty panels that don't specify shot size, camera move, or duration have decided nothing.

Principle 2: Duration sets panel count, not the story. A thirty-second spot holds about twelve shots at commercial pacing and about six at narrative pacing. Writers board the story they want and discover in the edit that it runs fifty seconds. Budget the seconds first and the story has to fit a real container, which is the constraint that makes it sharp.

Principle 3: The cut carries more information than the panel. Most video problems live at the boundary between two panels: the subject jumps to the other side of the frame, the eyeline doesn't match, the action restarts instead of continuing. A board that describes panel interiors and ignores the cut points hands every one of those decisions to whoever is cutting, who will guess.

Principle 4: Sound is half the board, and the muted version is the real one. Any board without an audio column per panel is a comic strip. At the same time, most social video is first seen with sound off, so the picture and on-screen text have to deliver the message alone. Both passes have to work, and they are different passes.

Principle 5: A panel that changes nothing is not a panel. If the viewer knows and feels exactly what they did one panel ago, that is a held shot, a longer duration on the previous panel, not a new one. Applied to a first draft, it is the fastest way to find panels to cut.

Layer 2: Systematic Approach

Step 1: Lock the deliverable spec before you draw anything. Write down six values: exact duration in seconds, aspect ratio or ratios, where it plays, whether sound is on by default, what the final frame is (logo, call to action, product), and who executes it (you, a crew, an editor, a generative tool). Every later decision depends on these, and a board built before they exist gets rebuilt.

Step 2: Write the beat spine in one line per beat. Four to six beats for thirty seconds: hook, setup, turn, payoff, end card. Name the turn explicitly, the moment the viewer's understanding changes. If you can't name it, the piece has no turn and no amount of boarding will add one.

Step 3: Assign seconds to beats so they total the exact duration. Subtract the end card first, usually 2 to 3 seconds. For social, the hook gets no more than the first 2 seconds. Write the numbers down: 2 hook, 8 setup, 6 turn, 11 payoff, 3 end card. A beat that can't survive its allocation gets cut at this stage, which costs a line of text instead of a shoot day.

Step 4: Panelize inside the time budget. Divide each beat's seconds by your target average shot length: roughly 2 seconds for ads and social, 4 to 6 for narrative. That gives the panel count per beat before you invent content. Eleven payoff seconds at commercial pacing is five or six panels, not nine.

Step 5: Spec every panel on eight fields. Panel number; shot size and subject distance (wide, medium, close, insert); camera position and any move, plus what motivates the move; subject action at cut-in and at cut-out; audio (dialogue or voiceover line, music cue, effects); on-screen text and where it sits; duration in seconds; and a one-sentence purpose line. Missing fields are the questions you will be asked later, in a text message, on the day.

Step 6: Run the continuity pass across the cut points, not down the panels. Read panel pairs. Check screen direction (a subject moving left to right keeps moving left to right across the cut), eyeline, light direction, props and wardrobe, and whether action continues or restarts. Fix by changing camera position, not by adding a panel.

Step 7: Run the muted pass and the blind pass. Cover the audio column and read the board: does the message land from image and on-screen text alone? Then cover the images and read only audio and purpose lines: does it still make sense? Anything that fails both passes is decoration.

Step 8: Flag everything that costs money or time. Per panel, mark location, cast count, wardrobe, props, animation or visual effects, and anything needing permission or a permit. Total the flags. This is the moment a board becomes a budget conversation instead of a surprise.

Step 9: Read the board against a clock before anyone commits. Out loud, at pace, timer running, one panel at a time at its stated duration. Real boards run long. Cut to the target duration here, on paper, where cutting is free.

Layer 3: Force Multipliers

Multiplier 1: The one-sentence purpose line per panel. Write what the viewer knows or feels after this panel that they didn't before. Panels that can't earn a distinct sentence are duplicates. Boards with purpose lines come out shorter, and review conversations shift from "I like this shot" to "this panel does nothing," which is the conversation you want.

Multiplier 2: Spec the cut, not just the frame. For each boundary, write the out-frame state and the in-frame state in a few words ("exits frame right, hand still raised" into "enters frame left, hand lowering"). This single habit removes most continuity notes in the edit, because the person cutting is matching a written state instead of guessing intent.

Multiplier 3: Board the last frame first. Decide the end card and the final image before the hook. Everything upstream then has a target to arrive at, and you stop writing pieces that build to nothing. Working backwards from the payoff also exposes setup you don't need.

Multiplier 4: Fix the eight fields as a template and never vary them. Same columns, same order, every board. Consistency makes an incomplete board visible at a glance, and it lets someone else fill panels in your absence without an explanation of your format.

Layer 4: Success Metrics

Positive indicators:

• Someone who was not in the concept meeting produces a shot list from the board without asking a question
• The timed read-through lands within 2 seconds of the target duration
• The board reads correctly with the audio column covered
• Review notes are about sequence and timing rather than about drawing quality
• Panels get traded during review instead of added

Failure signals:

• Every panel reads "wide shot of the subject doing X" (no move, no duration, no purpose: the board is illustration, not a plan)
• The timed read-through runs 30 percent or more over target (beats were never given a second budget, so the overrun was built in at step 2)
• The crew or editor asks who is standing where, or which direction someone is walking (screen direction was never specified, so continuity is being invented on the day)
• Reviewers keep asking to add one more panel and nothing comes out (no purpose lines, so nothing can be judged redundant)
• The first cut feels fine muted but incoherent with sound on, or the reverse (only one of the two passes was run)

Layer 5: Implementation Guidance

For solo operators vs. teams: A solo operator can drop the drawings and keep the fields, a text board with shot size, move, action, audio, duration, and purpose is fully sufficient when the same person shoots and cuts it. What a solo operator must not drop is step 9, the timed read, because there is no second pair of eyes to catch a forty-second thirty-second spot. Teams need the opposite discipline: one owner of the board file, one version, and flags totalled before the board circulates, or three people will spec the same panel three ways.

Common mistake: Drawing before locking duration and aspect ratio. The second most common is boarding the hook last, after the energy is spent, which is why so many spots open with a logo instead of a reason to keep watching.

Edge case: no duration given ("just make it short"). Pick one, write it at the top of the board, and say so when you send it: "boarded to 30 seconds, tell me if you want 15." Never board without a number. A 15-second version is not a trimmed 30, it's a different beat spine with the setup removed.

Edge case: the same piece has to ship 16:9 and 9:16. Board vertical native and mark which panels break in crop, usually anything with two subjects side by side or text near an edge. Do not plan to crop a wide master and hope. Where a panel can't work in both, board the alternate framing as a sub-panel, 4a and 4b, rather than compromising on one shot that serves neither.

Edge case: the board is feeding a generative video tool. Each panel needs a self-contained prompt-ready description, because the tool has no memory of panel 3 when it renders panel 4. Carry the consistency anchors forward explicitly in every panel: same subject description, same wardrobe, same light direction, same location phrasing, word for word. Also expect to shorten durations, since most tools produce usable motion in short clips, so plan the cut points around what the tool can actually hold.

Edge case: a dialogue scene with little visual variety. Panelize on who holds the turn in the conversation, not on each line. One panel per shift in who has power or information, with lines grouped underneath. Boarding line by line produces a board of alternating close-ups that decides nothing.

Edge case: a late request to add a panel. Trade, never add. The duration is fixed, so ask which panel it replaces. If the answer is "none," the answer is no, and the timed read-through is the evidence.

Where this doesn't apply

This is for planned, time-bound visual content where a spec exists. It is the wrong tool for unscripted and documentary capture, where you cannot board what hasn't happened yet, use a shot wish list and a coverage checklist instead. It also does nothing for live events. And it will not rescue a weak idea: a fully specified board makes a bad concept fail faster and cheaper, on paper instead of on a shoot day, which is worth real money but is not the same as making the idea good.

Worked example: 30-second spot for a mobile pet-grooming service

Spec (step 1): 30 seconds exactly. 9:16 native, 16:9 crop needed. Plays on social, sound off by default. Final frame is the booking link. Executed by one operator with a phone and a gimbal.

Beat spine and time budget (steps 2 and 3): Hook 2s (dog hiding behind a chair). Setup 8s (the usual version: crate, car, waiting room). Turn 6s (the van pulls up at the house). Payoff 11s (grooming happens in the driveway, owner stays in the kitchen). End card 3s.

Panels 1 to 3 of 13 (step 5):

Screen direction is set in panel 2 (left to right) and held through the setup beat, then deliberately reversed in the turn beat when the van arrives, so the reversal reads as the change rather than as a mistake. Total setup beat: 8 seconds across four panels. Production flags: one location plus a parking lot, one cast member, one dog, no permits, no effects.