How to write a comic script

What a comic script is for

A screenplay describes what a camera records over time. A comic script describes a fixed number of still images and the words printed on them. That difference drives everything: you are not writing action, you are choosing moments — the single instant of each beat that carries the most information.

The script exists so that whoever draws the page knows exactly what to draw, and so that whoever letters it knows exactly what goes in each balloon. If you are working alone with AI, the reader is the generator and yourself two weeks later — which, if anything, raises the bar for precision.

The standard structure

Almost every comic script, in any house style, nests the same four levels:

• Page — numbered. One page is one unit of layout and one page turn.
• Panel — numbered within the page. One panel is one image.
• Description — prose telling the artist what the image shows.
• Dialogue and captions — labelled by speaker, in reading order.

Written out, that looks like this:

PAGE FOUR (five panels)

PANEL 1
Wide shot. The rooftop at night, rain coming down hard.
MIRA stands at the very edge, back to us, cape soaked
and heavy. City lights blur below her.

CAPTION: She had been up here eleven minutes.

PANEL 2
Close on Mira's face, three-quarter view. Rain on her
jaw. She is not crying. She looks decided.

MIRA: I'm not going back.

PANEL 3
Reverse angle, low. The stairwell door behind her opens.
A wedge of yellow light across the wet concrete.

SFX: KRRRK

PANEL 4
Two-shot. DET. OKONKWO in the doorway, coat over his
head, one hand raised. Mira has not turned around.

OKONKWO: Nobody's making you.
OKONKWO: But it's a long walk down and I've got the car.

Panel five is left off deliberately — the last panel on a page is the one the reader sits on before turning, so it earns more thought than the rest.

Writing the panel description

A good description answers three questions and stops: who is in frame, how close are we, and what is the single thing happening. Everything else is either a constraint the story genuinely requires or clutter.

The most common mistake is writing motion. "Mira leaps across the gap and lands hard" is two images, and an artist has to pick one — so pick it yourself. Choose the instant that implies the rest: mid-air with the far ledge still out of reach, or the landing with the jump written in her posture. A still image can only ever show one moment, and deciding which one is the writing.

The second most common mistake is describing what cannot be drawn. "She remembers her sister's warning" is invisible. Either it becomes a caption, a flashback panel, or an expression — but it has to become something a reader can see.

How much detail is right

Two schools exist, and both work. Full script specifies every panel in detail before any art exists — precise, and the norm when a writer and artist are different people. Plot-first gives the artist a page-by-page summary of events and lets them break it into panels, with dialogue written afterwards over finished art.

Working with AI, full script is almost always the right choice, because the generator has no judgement to fill gaps with. But keep descriptions to what is visible. Details like "she is anxious about the meeting" do not render; "her hands are wrapped tight around the cup" does.

Dialogue in a script

Comics dialogue is compressed. A balloon that runs past roughly twenty-five words starts to crowd the art, and a panel carrying more than two balloons usually wants to be two panels. Reading this constraint as a limitation is a mistake — it is what keeps comics dialogue sharp.

A few conventions worth following:

• Label every line with the speaker's name, even when it is obvious.
• Write multiple balloons from the same character as separate lines, not one long block — that is how they will be lettered.
• List dialogue in the order it should be read, since that order determines balloon placement.
• Mark captions, off-panel dialogue, and sound effects distinctly (CAPTION, OFF, SFX).

Balloon order and placement are a layout problem as much as a writing one — see comic panel layout and reading order for how the two interact, and the guide to speech bubbles for shapes and tails.

Scripting for vertical scroll

Webtoons have no pages, so the page level of the structure disappears and is replaced by episode and, informally, by scroll beats. What matters instead is where you place the deliberate gaps — a large vertical space is the only page turn a webtoon has, and it is where you hide a reveal.

Episodes also end differently. A page ending invites a turn; an episode ending has to survive a week of waiting. The last beat of a webtoon episode carries much more weight than the last panel of a page.

From script to finished pages

A script is not the slow part. The slow part is that every panel description has to produce an image with the same characters in it — which is where most AI comic attempts collapse, since each generation reinterprets the cast from scratch.

In YarnSaga you define each character once and reference it in panel descriptions, so the script maps onto generated pages directly: pick the page layout, write each panel, and the same faces appear throughout. The dialogue you wrote stays dialogue — placed in balloons on their own layer rather than baked into the art.