Voice Criteria
The three character fields — Speech Pattern, Deflection Mode, and Tell — that keep every character's dialogue and behavior distinct from every other.
What is Voice Criteria?
One of the most common failures in AI-generated fiction — and in a great deal of human-written fiction — is the problem of character voice collapse. Every character begins to sound like the same person: the narrator. Dialogue becomes interchangeable. A reader could swap the attribution on two characters' lines and nothing would feel wrong, because nothing about the way they speak is distinctive enough to catch.
Voice Criteria exists to prevent this.
Three fields, applied to every named character in a story, that define not just how they speak but how they break under pressure. Because the moments that matter most in fiction — confrontations, revelations, confessions — are the moments of pressure. And pressure is when a character's voice becomes most distinctly itself, or most distinctly collapses.
The three fields
Speech Pattern defines the normal texture of a character's voice: their rhythm, their sentence length, their diction, their verbal tics, and — crucially — what they never say aloud. A character who speaks in long, subordinate-clause-heavy sentences sounds different from one who speaks in short declaratives. A character who uses academic vocabulary inhabits a different register from one who uses the plain language of manual work. A character who trails off, who asks questions instead of making statements, who repeats the last word of what someone says before responding — each of these patterns is distinctive and consistent.
Deflection Mode defines what a character does when they are asked something they do not want to answer. This is different from lying. Deflection is what happens before lying becomes necessary — the evasion strategy the character reaches for automatically. One character makes a joke. Another pivots to procedure: "I'll need to check the records on that." Another goes very still and waits for the question to pass. Another turns the question back: "Why do you ask?" Each deflection style is characteristic and reveals something about the person using it.
Tell defines the physical or behavioral habit that appears when a character is lying, afraid, or otherwise under stress. It replaces the authorial gesture of naming the emotion ("she was frightened") with a specific, observable behavior that the reader can recognize and track. A character who corrects small factual errors when she is scared. A character who goes very still. A character who becomes suddenly, compulsively tidy. A character whose speech slows and whose sentences become shorter. The Tell is the craft signal that something is wrong — legible to the attentive reader without being stated.
Why all three matter together
Speech Pattern defines the baseline. Deflection Mode defines the response to social pressure. Tell defines the response to emotional pressure. A character with all three defined is a character who exists in three registers — normal, socially stressed, and emotionally stressed — and sounds different in each one. That difference is what makes them feel like a real person rather than a function of the plot.
How Bespoke Books uses it
In the Bespoke Books Composer, Voice Criteria appears in the Characters tab for every named living character. All three fields are required. The generation pipeline injects them into every chapter beat header where the character appears: Speech Pattern and Deflection Mode feed into dialogue generation, Tell feeds into action and interiority beats. Characters who share similar Voice Criteria will begin to sound alike — the fields are specifically designed to create maximum differentiation between characters.