Storyboarding Before You Generate Anything

When a shot costs a crew, a location and a day, nobody generates it speculatively. The cost imposes planning. Generation removes that cost, and with it the discipline that came free — so the natural workflow becomes generating interesting images and working out afterwards what to do with them.

The result is a folder of individually good clips that do not cut together: inconsistent lighting, mismatched framing, no through-line, and several shots doing the same job. This guide covers a shot list format that fits generated production, what planning tells you about which shots to generate at all, and why the boards should stay rough.

What Planning Actually Saves

The obvious saving is time spent generating shots you do not use. That is real and it is the smaller half.

The larger saving is attempts. Generation costs are dominated by iteration — the fifteen tries before an acceptable result — and every one of those spent on an unnecessary shot is pure waste. A list that removes four shots removes forty attempts.

The second saving is discovering, before you start, which shots require continuity. Two shots of the same character need that character to match. Two shots in the same room need the room to match. These are the shots generation handles worst, and knowing which they are lets you either film them, restructure to avoid them, or accept the difficulty deliberately rather than discovering it at attempt thirty. What can actually be carried between generations is covered in the continuity techniques and what each one controls.

The third is the most valuable and least visible: knowing what each shot is for. A shot with a stated job can be judged. A shot that merely looks good cannot, which is why unplanned generation tends to accumulate rather than resolve.

A Shot List That Fits Generated Work

Work from the script, in order, one row per shot.

The continuity and source columns are the two that specifically suit generated production. Filling them in order tends to produce the same realisation each time: the shots that need to match are the shots that should not be generated.

Draw Roughly, or Not at All

Boards get abandoned because people try to make them look good.

The purpose of a frame is to settle three things: what is in shot, how big it is, and whether anything moves. A box with a stick figure and an arrow settles all three. A carefully rendered drawing settles the same three and takes twenty times longer.

There is also a specific risk in detailed boards for generated work. A polished frame sets an expectation about the exact image, and generation will not match it. A rough one communicates the intent — a close-up of hands, from the left, static — which is what you actually need to evaluate the result against.

If drawing is genuinely a barrier, skip the pictures. A written shot list with the fields above does most of the work, and a hybrid where only the complicated shots get sketched is often the best use of the hour.

Deciding What Not to Generate

Once the list exists, go through it once with a single question per row: does this shot have to match anything, or carry a factual claim?

Rows that need to match another shot are the hard ones. Rows that assert something happened are the ones where quality is not the only concern, because the audience is being asked to believe something — the test for when that requires telling them is in disclosing AI-generated media to your audience.

Everything else — texture, atmosphere, abstract illustration, establishing mood — is where generation is comfortably strong, and those rows can be marked with confidence.

There is a third option that gets forgotten in the choice between generating and filming: drawing. Where a row needs to show a structure, a process, or a relationship rather than a place or a person, an illustrated treatment is usually clearer than either, and it has no continuity problem at all because you control every element. The way such a sequence is built and paced is covered in structuring a whiteboard explainer that holds attention.

The Three Rows That Usually Get Cut

Filling in the job column honestly tends to expose the same three kinds of redundant shot, and recognising them early is most of the value.

The second establishing shot. A sequence often opens with a wide, and then another wide from a different angle, because both looked good. The second one establishes nothing that the first did not, and it delays the point at which the video says something.

The illustration of a thing already said. A shot whose job is to show what the narration is describing at that moment, where the description was already clear. This is the most common generated shot and the one with the weakest justification — the picture is agreeing rather than adding.

The transition that transitions nothing. An abstract clip between two sections that are already clearly separated by the narration. It reads as a pause in the argument rather than as a bridge.

Each of these survives an unplanned workflow easily, because generating them is quick and they look fine in isolation. They only reveal themselves as unnecessary when the list forces you to write down what they are for.

The reverse case is worth watching too. A row where the job column says something real but no shot exists for it is a gap, and gaps found on paper are cheap. Found in the edit, they mean going back to generation with the rest of the sequence already built around a hole.

The Board Survives Into the Edit

A shot list is not discarded once shooting or generating is done — it is the structure of the edit.

Numbered rows become numbered files, which makes an assembly edit close to mechanical. The job column tells you, when a shot has come out weak, whether the sequence still works without it or whether something else has to cover that function. And the duration column keeps you honest about generated clips, which look best short and tend to be held too long because they were expensive to obtain.

It also gives you a basis for judging a near-miss. A generated shot that is not what you drew but does the stated job is a success; one that looks striking and does a job already covered two shots earlier is not. Without the list, both look like wins.

The prompts themselves come straight out of the rows: subject from the script line, framing from the camera column, duration from the duration column. That mapping is deliberate, and the structure it feeds is set out in writing prompts that produce usable video.

Next Steps

Take the last generated sequence you made and reverse-engineer its shot list — one row per shot, with the job column filled in honestly. Count the rows where you cannot name a job. That count is what planning would have saved.

Then board your next project before opening any tool, even roughly, even as plain text. The hour it takes is reliably less than the time lost generating shots that turn out to have no place. Neighbouring articles are listed on the generation hub.