How to Write Dialogue in Comics and Graphic Novels

Balloon, caption, thought

A speech balloon is the character talking, and it should be treated as a shot size, not as a caption with a wiggly border. One balloon carries one move: a question, a dodge, a fact that changes the next panel. The words inside are what we hear. They are not a second copy of what the drawing already shows. If Pilar is clearly burning a letter, the balloon should not say she is burning the letter. It should say the thing the burning does not say, or it should be absent. Lettering takes space the art needed. Every word is ink you stole from a face, a hand, or a background that was doing story. Write the balloon after you know what the panel shows, and be prepared to lose the line when the picture got there first. A balloon that wins a fight against its own panel is a balloon the page cannot afford.

A caption is not a balloon that you forgot to give a tail. It is narration, time, place, or a voice that is not coming out of a mouth in that panel. Captions usually sit in a box, and they do not point. Use them to move us, three hours later or the same kitchen seen from a new hour, and not to explain the plot the balloons are already acting out. A caption that repeats the balloon is a writer who does not trust the reader to read. A caption that repeats the art is a writer who does not trust the artist. If you are both, you still do not get to say it twice. Distinct caption boxes can separate two narrators, but only if you set the rule early and keep it. A sudden rectangular box late in the book that means this is irony now will be read as a location change. Decide what the box is for, then let it be boringly reliable. Reliability is how a caption stays invisible enough to work.

Thought balloons, the cloud shapes with a trail of little bubbles, are a special effect. Use them rarely, and suspect them even then. A thought balloon speaks the quiet part at the same volume as dialogue, which flattens the gap between what a character says and what they want. Modern graphic novels often skip the cloud entirely. They put a private line in a caption with a box you have already taught us to read as interior, or they trust the face. If you use a thought balloon, spend it on a thought the picture cannot draw and the dialogue must not say. One in a scene is already a lot. A page of clouds is a radio play pasted over the art. Do not combine a thought balloon, a caption, and a speech balloon that all confess the same feeling. Pick the container that belongs to the kind of information it is. Speech is heard. Captions are read as story. Clouds are a last resort, not a personality you owe every brooding character.

PAGE 3

PANEL 1
Pilar at the kettle. Night. The porch light is off in the window behind her.

CAPTION: Night. The kettle has opinions.
PILAR: I was home.

PANEL 2
Benno in the doorway. One balloon, tail to his mouth.

BENNO: Then why is the light out?

PANEL 3
No balloon. Pilar’s hand stays on the cold kettle.

The tail is the pronoun

The tail is a pronoun. It points at the speaker’s head or mouth, not at their belt, not at the person they are accusing, not into empty air because the composition was crowded. A tail aimed wrong gives the line to the wrong body, and the scene changes meaning without a single word changing. In a crowd, the tail is often the only way we know who spoke. Separate the bodies enough that two tails cannot be confused, or crop the panel so only one mouth is a plausible owner. Do not write the speaker’s name inside the balloon to compensate for a bad tail. Names inside balloons are a transcript. They are what you do when the picture failed, and they eat the space the picture needed. If you are scripting before layouts exist, write the speaker’s name in the script, outside the balloon text, and add a note when the tail will be ambiguous. The letterer and the artist cannot read your mind. The tail is the mind, on the page, and it has to be obvious at a glance.

A character can have two balloons in one panel, but the reading order between them has to be obvious, usually higher then lower, or a connecting bridge the letterer will understand. In the script, number them or mark the second as connected. Do not assume that the longer one is first. Readers do not measure. They follow the path their eye was trained on. If both balloons point at the same person, we should still know which sentence happens first, because dialogue is time. Two unlinked balloons at the same height are a tie, and ties get read in the wrong order by half the room. Break the tie in the layout note. If you cannot break the tie, you have two beats and you wanted one panel. Give the second beat the next panel. Silence between panels is cheaper than a confused tail, and it is often the better cut. A connected balloon is one sentence with a breath. Two balloons are two moves. Know which one you wrote before you ask the letterer to guess.

PANEL 2
Pilar and Benno stand shoulder to shoulder.

BALLOON, tail drifting between them: I was home.
BENNO, printed inside the same balloon just in case: Then why is the light out?

PANEL 2
Pilar left, Benno in the doorway, a gap of wall between them.

PILAR (tail to her mouth): I was home.
BENNO (tail to his mouth, balloon lower): Then why is the light out?

Reading order is the cut

In a left-to-right comic, the reader enters a panel from the upper left and moves across, then down. The first balloon should be the one they hit first, which means it lives in that path, not merely first in your script. If the reply sits above the question, the scene plays backwards, and the joke arrives before the setup no matter how you numbered the balloons in the margin. Right-to-left books flip the entry. A manga page, or a translated edition that keeps that direction, does not owe you the Western path. If you are writing for that direction, decide it on purpose and place the first balloon where that eye actually enters. The failure in either tradition is the same: a writer who believes script order can overrule the page. It cannot. Layout is the edit. Your numbering is a proposal the art will betray if the balloons are parked in the wrong corner. Go look at the corner before you defend the sentence.

Large balloons on the right of a Western panel will be read first if they are also the highest mass on the page, because the eye hits bulk before it hits your good intentions. A tiny question tucked at the lower left cannot set up a huge answer floating over the character’s hair. If the art director moves your second balloon up to balance the panel, your scene may now run backwards, and that is not their failure to respect literature. It is the page doing its job. Write balloons short enough, and ordered clearly enough, that a layout pass cannot accidentally invert the plot. When you letter from a finished layout, read the page with your finger, in the order a stranger’s eye will travel, before you fall in love with the wording. If your finger hits the car before it hits the question about the light, swap the balloons or cut one. The script’s numbers are not a contract the retina signed.

PANEL 4
Left-to-right page. Benno’s balloon is upper left, tail to him. Pilar’s reply is lower and to the right, tail to her.

BENNO: Then why is the light out?
PILAR: Because of the car.

Do not swap those positions to “balance” the panel. The reply is not allowed to sit where the eye enters.

The word count that smothers a panel

Roughly twenty to twenty-five words is where a balloon starts to suffocate a panel. That limit is per balloon, not a budget you can spend all in one cloud and call efficient. Two balloons of twenty words are already a poster. A panel that also holds a figure, a background, and a sound effect does not have forty words of spare air. Count the words. If you are over, the first cut is any clause that restates the picture. The second cut is the explanation of the joke. The third is the adjective you added because the face was not drawn yet. When the face is drawn, the adjective will be insulting. Captions count toward the panel’s patience even when they are not the balloon. Sound effects count as lettering the eye has to parse even when they are not dialogue. A panel with a caption, a balloon, and a big sound effect is three reads. Most panels can afford one or two. Three is a crowd scene made of text, and the drawing will lose.

If a speech is too long for one balloon, do not shrink the lettering until it fits. Shrinking type is the comic version of narrowing the screenplay margins: it hides a length problem inside a production trick, and it punishes the reader who is already doing the work of looking. Break the speech across panels when the break itself is a beat, a face reacting, a hand stopping, a door that has not opened yet. If the break does not change anything, you do not need the extra panel and you do not need the extra words. Cut. A balloon split by a panel border should still be readable in order, with tails that keep the speaker, and the second half should not repeat the first half for people who might have missed it. They did not miss it. They are holding the previous panel in their eye. Trust the next panel on the same page more than you trust a recap, and trust a shorter line more than either.

PILAR: I am standing here with my hand on the kettle, which is cold, which proves I was not home the way I said I was, and I am sorry you had to see the car in the drive because I know what that looks like to you.

PILAR: The kettle’s cold.

PANEL next. No caption.
PILAR: I did not want the light on.

PANEL next. Still no recap.
BENNO: I saw the car.

Continuity across panels

A conversation that continues has to look like a continuation. Do not restart every panel by having the character say the other person’s name as if we have just tuned in. Do not answer a question in a balloon floating over a brand-new location unless a caption has moved us, or the art has moved us so clearly that a caption would be redundant. If the speaker changes, the tail changes, and the new tail should be unmistakable. If the same speaker continues across a gutter, keep their position on the panel consistent enough that we are not re-solving the scene. Letterers sometimes join balloons that continue a single sentence. If you need that, mark it in the script as one sentence broken, not as two speeches. Two speeches can be contradicted. A broken sentence is one thought with a breath in it. Know which one you wrote, and do not make the second half introduce a new lie the first half did not earn. Continuity is not repetition. It is the reader not having to re-roll the scene every time the gutter arrives.

The page turn is a cut you do not fully control, because the reader has to choose to take it, and the right-hand page is the last thing they see before the choice. Do not put the answer to a question in the first balloon of the next page if the question was a whisper in the bottom corner of this one, buried under a caption. Give the turn a button: a face, a short line, a question that fits in one balloon. After the turn, start clean enough that a reader who is going at speed can land. The last balloon before the turn should not depend on a trick that only works if you remember a tail from three panels back. Say the necessary noun again if the turn would otherwise strand it. This is not the same as restarting the conversation every panel. It is acknowledging that a page turn is a small amnesia. Write for the amnesia. Then make the next page worth the reach. A turn that exists only because the script ran out of room is a turn that will be resented.

PANEL 5
Same kitchen, Pilar still left of frame. One balloon, tail to her mouth.

PILAR: I was home.

PANEL 6
She has not crossed. The balloon sits lower than the last one would have, still hers. No name inside it.

PILAR: I just did not want the light on.

PANEL 7
Page turn. Benno’s face, no recap caption.

BENNO: The car was in the drive.

Voice without a phonetic costume

Phonetic spelling is a worse idea in a balloon than it is in prose, and it is already a bad idea in prose. The letterer has to draw every dropped letter, the balloon swells, and the reader has to solve a puzzle at the speed of a panel instead of hearing a person. Dialect and voice live in syntax, rhythm, and word choice: what someone bothers to say, which noun they reach for, how long a sentence they trust. They do not live in comic misspellings of a community’s speech. If you are writing a real community’s language, research it, hire a reader from that community, and let the lettering stay legible. A balloon is a terrible place to perform an accent you do not speak. The note at the top of this page is there because phonetic dialect has a long history of being used as a joke at the speaker’s expense. Restraint will not make the voice thinner. It will make it readable, which is the whole job of lettering, and readability is not the same thing as sanding everybody down into the same announcer.

You can still hear two people if their balloons want different things and waste different words. Pilar answers with objects. Benno answers with locations. One of them uses a short sentence when they are lying and a longer one when they are bargaining. None of that requires a misspelled vowel. If a character’s language includes a word the other character does not use, write the word properly and let the tail do the rest. Code-switching belongs in real words, not in a cloud of apostrophes. And if you need a language the page is not printing in full, a caption or a brief note in the script can say the line is in that language, while the balloon carries what you have chosen the reader to understand. Do not invent a fake orthography to gesture at a language you have not written. The artist can show who is not being understood. The balloon’s job is to be read. When it becomes a stunt, the panel stops being a scene and starts being a demonstration of the writer’s ear, which is the wrong person to star.