How to Write Dialogue That Sounds Real
Record an actual conversation and type it up exactly. It will be unreadable. Dialogue that sounds real is a performance of speech, not a recording of it, and the difference is craft you can learn.
Why a transcript sounds fake
Genuine speech is mostly maintenance. People stall, repeat themselves, restart sentences, confirm they were heard, and spend a surprising amount of breath on hello and goodbye. In a room none of that registers, because the ear is patient and the body is filling in the rest. On the page it registers as every line, and the reader starts skimming.
So the reader's sense of "this is how people talk" comes from compression. You keep the texture of speech, the interruptions, the sideways answers, the sentence someone never finishes, and you throw away the parts that carry no charge. The oldest working rule still holds: enter a conversation as late as you can and leave it as early as you can. If the scene opens with two characters saying each other's names, you started a page too soon.
People talk around what they want
Most bad dialogue is bad because everyone is honest. Characters explain their feelings, state their motives, and hand each other the information the plot needs. Nobody does this. In real conversation there are always two things running at once: what a character wants, and what they are willing to be seen wanting.
Dialogue lives in that gap. "Did you eat?" can mean I love you, or I've been waiting three hours, or I know where you were. The line on the page is the smaller, safer version of the thing underneath, and the reader does the arithmetic. That act of solving is the pleasure.
A practical way in: before a scene, write one sentence for each character stating what they want from the other person, and one sentence stating what they would rather die than say. Then make sure neither of them says either sentence. What survives is dialogue with pressure behind it.
Give each character a different mouth
Distinct voices rarely come from catchphrases or verbal tics, which wear out by the third appearance. They come from the way a person's history shapes their sentences:
- Vocabulary and register. A paramedic, a priest and a teenager name the same object three different ways. Occupation and education do more work than any accent.
- Rhythm and length. Some people talk in long qualified paragraphs. Some answer in four words. The shape of a speech is visible on the page before a word is read.
- Behaviour under pressure. One character gets formal when frightened, another gets profane, a third goes quiet and asks questions. Stress is where voice shows.
- What they never say. A character who cannot say sorry, or cannot say the dead man's name, is characterised by the hole in their speech.
The test is mechanical. Cover the character cues or the dialogue tags and read the exchange. If you cannot tell who is speaking, the voices are still yours rather than theirs.
Dialect without phonetic spelling
The instinct with a regional or period voice is to respell it: dropped letters, apostrophes everywhere, vowels rendered as the writer imagines them. It backfires twice. It slows reading to a crawl, and because the narrator's own words stay in standard spelling, it quietly marks the speaker as the odd one, which is how caricature happens.
Accent lives in grammar more than in sound. Move it into:
- Word order. Where the emphasis falls in a sentence is regional, and it survives standard spelling intact.
- Idiom and reference. The phrase someone reaches for, the thing they compare the weather to, what they call the evening meal.
- Contraction and formality. Whether a character says "cannot", "can't" or "I couldn't say", and whether they use your name in every sentence.
- A very light spelling touch, early. One or two accepted forms in the first speech, then stop. The reader has cast the voice by then and will keep hearing it for free.
Period voice follows the same discipline, with one extra hazard: idiom is dated. A phrase that feels timeless often arrived in the 1970s. That belongs to the research pass, not the ear.
Tags, beats, and the space around the words
The machinery around dialogue is where a lot of good lines get spoiled.
- "Said" is invisible, so use it. Readers stop seeing it after two pages. Hunting synonyms puts a spotlight on the tag rather than the line.
- An adverb on a tag is usually a patch. "He said angrily" tells you the line failed to be angry. Fix the line.
- Action beats do two jobs. They control pace, and they let the body contradict the words. "Fine," she said, still holding the car keys.
- Silence is a line. A character who answers a question by changing the subject has answered it.
Screenwriters have a narrower version of the same problem, since a script has no tags at all. Everything in the list above collapses into character cues, sparse parentheticals and the action between speeches. See the screenplay format guide for what belongs in each element.
Your ear is the only reliable judge
Silent reading is a bad instrument for dialogue, because your inner voice cheats. It knows where the emphasis goes, it knows the joke, it pauses in the right places, and it does all of that whether the line earned it or not. Read the same page aloud and the repairs stop. Suddenly the sentence has no comfortable breath in it, two characters turn out to share a rhythm, and the clever exchange needs a tone of voice the words never supplied.
Hearing it in someone else's voice is better still. That is what a table read is for, and why theatre and television have done one before rehearsal for as long as either has existed. Pages that read beautifully in private have a way of dying at a table with actual voices in the room, and it is much cheaper to find that out early.
How Skarvia gives you the table read
In the Review tab there is a Table Read. You cast a voice for each speaking character and for the narrator, set a pace, and press play. Skarvia reads the scene back and highlights each line as it speaks it, so you follow the page while you listen and can see exactly where your attention slid off. In a novel or an A/V script the same tool is Read Aloud, with the narrator carrying the prose.
One detail is deliberate: the read skips parentheticals. If a line only works once you hear "(sarcastic)" in front of it, the line is not doing the work yet, and hearing it without the instruction is the fastest way to know. Next to it, Insights counts cues and words for every speaking character, which is how you catch the one who has quietly taken over the scene, and the one who has not spoken since page forty. Neither tool proposes a better line. Skarvia never writes a word of your dialogue; it hands you the room and the ear, and the rewrite stays yours.