Dialogue can work perfectly well on the page and fall apart the moment someone reads it aloud. We confirmed this while working on the Voctale episodes: lines that looked clean on screen turned wordy at the microphone, and sentences we had thought too clipped finally found their place. The ear does not judge the way the eye does. It moves faster, it has less patience, and it spots anything that sounds written.
We covered the format itself in Writing an audio story: what the format really changes. This time we are going one level down, to the individual line.
The eye forgives, the ear does not
When we read silently, we reread without noticing. We double back over a badly placed subordinate clause, rebuild the sentence, and the effort goes unregistered. In listening, that backward step does not exist: the voice keeps moving, and a sentence that needs two passes is a sentence lost.
Take a line written for the page:
"The notebook that my grandmother, who never wrote anything down, had filled during the winter of '44 was still up in the attic."
Typographically there is nothing wrong with it. Read aloud, the listener drops out at the second comma. Same information, redistributed:
"My grandmother never wrote anything down. That notebook, she filled it during the winter of '44. It stayed up in the attic."
Three breaths instead of one long hold. Nothing was lost, and the performer now knows where to place her voice.
That shift clears out three writing habits fairly quickly: long parentheticals in the middle of a line, pronouns whose referent sits far behind them, and chains of relative clauses. They are comforts of silent reading, not of listening.
Reading aloud is still the only revision that counts
There is no shortcut. Neither word count nor a readability score from a plugin will tell you whether dialogue works. You have to hear it, preferably in your own mouth before an actor's.
Our method is a short one. We record the scene on a phone microphone, without trying to perform it well, then listen back and mark only the places where we stumbled. A swallowed word, a breath in the wrong place, a hesitation over the intonation: almost every time, the problem is in the text rather than the reading. A line built for the ear sounds right even when it is badly delivered.
Recording does something less obvious too: it puts a clock on the scene. A page of tight dialogue runs to roughly a minute and a half of listening. When a scene you imagined as quick comes in over four minutes, the pace of the performance is not what is at fault.
Giving each character a way of speaking
In audio the listener cannot see who is talking. If two characters build their sentences the same way, they will be confused with each other, and no amount of direction will fully compensate.
Rhythm before vocabulary
The temptation is to separate characters by their word choice: one speaks in slang, the other formally. It registers, but it ages badly and slides quickly into caricature. Rhythm is quieter and more durable. A character who answers in short sentences, never using a conjunction, stands clearly apart from one who piles clauses up and corrects himself, even when both are using the same words.
— Did you tell her?
— No.
— Why not?
— Wrong moment.
Against:
— I meant to tell her, well, I'd started, and then I realised she was already on her way out, so, you know.
Two characters, no name tags, and nobody gets them mixed up.
Verbal tics, in very small doses
A filler word that belongs to one character – right, anyway, look – works if it comes back three or four times across an episode. Beyond that, the listener stops hearing it as characterisation and starts hearing it as the writer's habit. The rule we hold ourselves to: if the tic is noticeable in a silent read, it is already too frequent.
Punctuation as a score
On the page, punctuation answers to grammar. In dialogue written to be spoken it becomes a performance note, and you have to accept that it will drift from the standard.
A full stop often replaces a comma purely to force a halt. Ellipses, on the other hand, need watching: they ask the performer to suspend the line, which costs time and only registers when it is rare. The em dash does not register at all – whatever it signals has to be carried by the syntax, or it vanishes in the recording.
Italics raise the same problem. A word underlined visually has to be stressed some other way: through its position in the sentence, or through the line that precedes it. She said that and "she was the one who said it" are not the same piece of writing.
What you can cut without losing anything
Dialogue almost always improves when it is shortened, and the same elements keep turning up under the knife.
Greetings and preambles first: a scene begins after the hello. Then vocatives. In life we rarely say the name of the person we are speaking to, and dialogue that leans on them is admitting it needs to label the voices for the listener. Better to settle that once, cleanly, in the opening line, then trust the timbres.
Finally, anything that repeats information the narration has already given. It is the most common fault in audio, because we are afraid the listener has drifted. The fear is reasonable, the fix is not: one well-placed line of narration beats a character explaining to another what they both already know.
A simple test before recording
Before taking a scene into the studio, we have it read by someone who did not write it, with no context given, and we listen with our back to the screen. Two questions are enough afterwards: at what point did we lose track of who was speaking, and at what point did we want the scene to move on. What a read-through reveals in five minutes, a silent revision can miss for weeks.
The habit ends up spreading to everything else. Since we started working this way, our texts that were never meant for the voice have got shorter too – rereading is, after all, where a way of writing comes from, as we described in Finding your writing voice.
For more on written-to-be-spoken fiction, screenwriters remain the best teachers: the series we cover on Subway Press are full of textbook cases, and long-form dialogue work also comes up on allisonlister.com.
0 Comments