See how the Dialogue Coach reads a scene: the tags and beats that pull attention away from what your characters are saying, and why.
Below is a sample chapter page and the notes Dialogue Coach returned for it. Click a note to find its line; click a highlighted line to find its note.
Dialogue Coach
Selection of 770 words · 2 credits in the desk
Real output
The Salt Ledger · Chapter 3: Three Bells
Story notes: Romantasy (adult) · Third person limited, Tamsin · Past tense
The bells of the Tithe House were ringing when Tamsin Harrow reached the top of the stairs, and she felt a wave of dread wash over her. Three bells meant a binding. Three bells meant someone in the lower city was about to sign away something they could not afford to lose.
She saw the envoy before he saw her. He was standing at the long table with his back to the windows, and the grey light off the marsh made him look carved rather than born. Corin Vael was arrogant and cold, the kind of man who had never once been told no and would not know what to do with the word if he heard it.
“You’re late,” he drawled lazily.
“The tide was in,” Tamsin snapped angrily. “Some of us walk.”
“Some of you,” he purred, “should learn to read a tide table.”
She clenched her fists. She was furious. She made herself unclench them, one finger at a time, the way her mother had taught her to before a binding, because anger made the ink run.
The salt ink was already laid out on the table. She noticed that it had been poured into the old pewter well, the one with the crack along its lip, and she realized that someone had wanted her to see it. Salt ink did not forgive. A contract written in it held for exactly as long as both parties kept their word, and the moment one of them broke it, the letters burned white into the breaker’s skin. Tamsin had seen a man lose three fingers to a debt of eleven silver crowns. She had been nine.
“Sit,” Corin said.
“I’ll stand.”
“Then stand,” he chuckled darkly. “It makes no difference to the Court whether the binder’s knees are comfortable.”
She heard the scratch of a quill somewhere behind the screen at the far end of the room. A clerk, probably. The Court never sent an envoy without a witness, and never sent a witness it trusted.
Tamsin felt nervous as she pulled the tithe roll toward her. It was heavier than it looked, as they always were, the vellum thick with the names of every family in the lower city who owed the Court a year of service. She could see the columns of names blurring together. Brewers. Net-menders. A family of chandlers she had bought candles from all her life. She was sad about the chandlers.
“You don’t have to read all of it,” Corin said quietly. “Only the names marked in red.”
“I read all of it,” she hissed. “That’s the job.”
“The job,” he breathed, “is to bind what the Court asks you to bind.”
She looked up at him then. His eyes were the color of the marsh at dusk, and there was something in them she did not like, something that looked almost like pity. She realized he knew something she didn’t, and she felt suspicious.
She turned the page.
The red names began halfway down. Orrin Fisk. Della Mott. The Pell twins, who could not have been more than fourteen. Tamsin ran her finger down the column, steady, steady, until it stopped of its own accord, because her hand had read the name before her eyes would agree to.
Bram Harrow.
Her brother. Her brother, who was nineteen and stupid and kind, who still fed the gulls on the seawall with bread they could not spare, who had sworn to her on their mother’s grave that he had paid the Court in full last spring.
The room was very quiet. Tamsin felt shocked and betrayed and terrified all at once.
“There’s been a mistake,” she said.
“There has,” Corin agreed. He set a second sheet of vellum beside the roll, and she saw that it was already written in salt ink, the letters faintly smoking where the light touched them. “Your brother borrowed against his tithe. Twice. The second time, he signed for you.”
“He wouldn’t.”
“He did.” Corin’s voice was gentler now, and somehow that was worse. “The Court is willing to accept a substitute. It usually is.”
Tamsin understood then what the cracked inkwell was for. Her mother’s inkwell. The one she had sold to pay for the burial. Someone had bought it back, and carried it across the marsh, and set it out on this table so that she would know, before anyone said a word, exactly how long they had been planning this.
“Who?” she asked.
Corin smiled, and it was not a kind smile. “Sit down, Miss Harrow,” he said softly. “This is going to take a while.”
7 notes, 2 strengthsClick a note to find its line
Summary
This exchange moves with clean escalation and a strong power dynamic; Corin’s cool lines and the final soft threat land especially well. The main thing dulling the edge is the reliance on said-bookisms and adverb-heavy tags, plus one non-speech tag, which steer tone and occasionally miscue character. Leaning on simple said/asked, selective action beats, and letting the words carry menace will make the voices and tension you already have read sharper.
1Major
Said-bookism plus an adverb in the dialogue tag; this pattern recurs in roughly eight tags here (drawled lazily, snapped angrily, purred, chuckled darkly, hissed, breathed, said quietly, said softly).
Why it matters
Decorative tags pull attention off the words and start telling the reader how to hear them, which can flatten the tension you’ve built. In a charged two-hander, cleaner tags or none let the power in the lines do the work.
Try this
Audit this exchange and swap most decorative tags for either "said/asked" or a brief action beat, and drop tags altogether where turn-taking is crystal clear. Let Corin’s diction and Tamsin’s defiance carry tone without the tag coaching it.
2Minor
Tone skew from the verb choice; “purred” carries a sensual or feline quality that clashes with Corin’s established cold control.
Why it matters
Readers may misread the dynamic as flirtatious or camp when the scene is staking out menace and hierarchy, blurring genre signals at a key first impression.
Try this
Pressure-test each expressive tag against Corin’s persona and the mood of the moment. If the goal is controlled threat, let his restraint and the content of the jab create the chill.
3Minor
On-the-nose emotion telling in a beat right around dialogue.
Why it matters
Stated emotions can feel generic next to the vivid setup and can undercut the charge in the spoken lines; readers engage more when they infer feeling from behavior and word choice.
Try this
Lean on the physical detail you already have (fists, ink, breath, focus) and on what she chooses to say or not say to let anger read without naming it.
4Major
A non-speech verb used as a speech tag; one cannot literally chuckle words.
Why it matters
This trips the ear and can knock a reader out of the moment, especially when stakes are rising. It also makes Corin sound amused when the line is really about power.
Try this
Separate the sound from the speech so the action reads as action, or keep a neutral tag and place the chuckle as its own beat before or after the line.
5Strength
A line of dialogue encapsulates power and world order without explanation.
Why it matters
This kind of status move lets the reader feel the Court’s indifference and Corin’s leverage in-story, sharpening stakes and character without slowing the scene.
Try this
Keep threading these cool, precise assertions where Corin can establish rules or limits in his own words, and let Tamsin’s replies test or refuse them to keep the tension live.
6Minor
Exposition delivered in a clean block without resistance or interruption.
Why it matters
Because this is a gutting reveal, a smooth info-drop can feel mechanical; readers expect friction—deflection, denial, or a power move—to texture how truth lands.
Try this
Consider breaking this into shorter beats with a pause, an interruption, or a withheld piece that Tamsin has to pry out, so the exchange dramatizes discovery rather than reports it.
7Strength
Restraint in diction and cadence creates menace and closes the scene on a controlled turn.
Why it matters
Understatement here lands harder than threat would, promising the scope of what’s to come and deepening Corin’s voice.
Try this
Use this measured, economical tone at key pivots; when the plot escalates, letting the language cool can make the danger feel larger.
›The tool’s summary
This exchange moves with clean escalation and a strong power dynamic; Corin’s cool lines and the final soft threat land especially well. The main thing dulling the edge is the reliance on said-bookisms and adverb-heavy tags, plus one non-speech tag, which steer tone and occasionally miscue character. Leaning on simple said/asked, selective action beats, and letting the words carry menace will make the voices and tension you already have read sharper.
Note 1 of 7
1Major
Said-bookism plus an adverb in the dialogue tag; this pattern recurs in roughly eight tags here (drawled lazily, snapped angrily, purred, chuckled darkly, hissed, breathed, said quietly, said softly).
Why it matters
Decorative tags pull attention off the words and start telling the reader how to hear them, which can flatten the tension you’ve built. In a charged two-hander, cleaner tags or none let the power in the lines do the work.
Try this
Audit this exchange and swap most decorative tags for either "said/asked" or a brief action beat, and drop tags altogether where turn-taking is crystal clear. Let Corin’s diction and Tamsin’s defiance carry tone without the tag coaching it.
The notes this tool returned on an original sample passage, unedited. AI feedback can be wrong, and in the desk you can dismiss any note. Your own pages get notes of their own.
Dialogue tags are small, and they quietly decide whether a conversation feels natural. In the workspace, the Dialogue Coach reads your scene and finds the tags and beats that may be pulling attention away from what the characters are saying, along with voices that blur together and lines that say the subtext out loud. It explains the effect on a reader and suggests an approach. It does not replace your lines.
How dialogue tags work
A tag is the short attribution that tells the reader who is speaking: he said, she asked. Most readers skim past said and asked the way they skim past punctuation. That is why many editors recommend them as defaults: they identify the speaker without competing with the dialogue.
Trouble starts when the tag draws attention to itself:
Said-bookisms.He growled, she hissed, he opined. These sometimes work, but a page full of them feels busy.
Adverbs doing the dialogue’s job.“I’m fine,” she said sadly. If the line doesn’t carry the sadness, the adverb is covering for it.
Tags that can’t produce speech.“Fine,” she smiled. You can smile while speaking, but you can’t smile a word.
Unclear speakers. Long runs with no attribution, or three or more people talking, leave the reader counting back.
Beats on autopilot. A gesture on every line, or one that does no work, slows the exchange without adding character.
Action beats are a good alternative to tags: a gesture or a bit of business next to the line (She set the cup down. “Where are you going?”) identifies the speaker and adds character. Overusing them has its own cost, so the aim is variety and clarity, not a rule.
What the tool looks for
1Said-bookisms and adverb-propped tags that pull attention off the words.
2Tags that can’t produce speech (she smiled, he laughed).
3Over-tagging in a two-person exchange, and under-tagging where the reader loses track of who is speaking.
4Action beats on every line, and beats that characterize well.
5Voices that blur together, and on-the-nose lines that say the subtext out loud.
6Exchanges that only mark time, and punctuation that makes a beat read as a tag or the reverse.
7At least one strength, so you know what to keep.
How the tool decides
The tool sends the passage or chapter you choose, with your story notes, to an AI model with a rubric based on widely taught dialogue craft. It reads the dialogue and the tags and beats around it, and returns quoted notes in the order they appear, at least one of them a strength, with a summary of what the dialogue does well and the one habit to change first. If the passage has no dialogue, it says so rather than inventing notes.
It is not a grammar checker. It looks at punctuation only where it changes whether a line reads as a tag or a beat. False positives are common in some voices: a stylised fantasy or a comic narrator may use vivid tags on purpose, and a fast two-person exchange can safely run untagged for a long stretch. Decide what suits your book.
About the example, and your text
The examples on our tool pages are real notes the tools returned on original sample passages we wrote. They are stored with the page: viewing them, or clicking through them, sends nothing to an AI model.
When you run the tool on your own pages, it sees only what it needs: the passage or chapter you choose, your story notes and the titles.
We don’t use your writing to train AI models. The AI providers we use process your text to return results and, under their API terms, don’t use it to train their models.
Yes, in the workspace, on a selection up to 2,500 words or a whole chapter up to 8,000, on the Writer or Author plan. Every first subscription comes with a 7-day money-back guarantee. Compare plans.
Should I only ever use “said”?
No. “Said” is a reliable default, not a law. The question is whether the tag helps the reader or distracts them.
Are adverbs banned?
No. They are a signal worth checking: does the line already carry the feeling?
What about “he said, smiling”?
That construction implies two things happening at once, which is sometimes what you mean. A separate beat often reads cleaner. The tool will point it out when it seems to cost you something.
Does it fix punctuation in dialogue?
No. It looks at attribution and rhythm, and mentions punctuation only when it changes how a tag or beat reads.
Are the notes on this page real?
Yes. They are what the tool returned in a test run on an original sample passage we wrote, shown unedited and listed in the order their lines appear. Nothing on this page sends text to an AI model. Your own pages will get notes of their own.
Is my text used to train AI?
No. We don’t use your writing to train AI models. The AI providers we use process your text to return results and, under their API terms, don’t use it to train their models. Details are in our AI promise.