The research

Your style is measurable. So we measure it.

Every question Timbrel asks is tied to a real body of research — stylometry, Biber's analysis of how language varies, the psycholinguistics of personality, and the mechanics of how models actually choose words. Here is the science, in pictures.

A fingerprint woven from flowing strands of language — a voice made measurable

Built on a deep-research review of stylometry, psycholinguistics & how language models generate text.

The problem

AI writing is average — on purpose.

A model is trained to please the widest possible audience, then polished by raters who reward the safe and the fluent. So it slides toward the center of everything it has ever read. One giveaway: humans are lumpy. Our sentence lengths swing; a model settles into one comfortable length and stays there.

Many near-identical pale swirls with one vivid mint-and-teal swirl in sharp focus — a voice that isn't average
Every default draft is one of the pale ones. Yours is the one in focus.
YOU THE MODEL
Each bar is one sentence. Human writing swings — the field calls it burstiness. Default machine writing flatlines.
Stylometry

Your fingerprint is made of small words.

The signal that gives a writer away doesn't live in the big content words. It lives in the glue you stopped noticing in second grade — the, of, that, while, upon, by. Stable across every subject you'll ever touch, as individual as handwriting.

In 1963, Mosteller and Wallace settled the disputed Federalist Papers with exactly this. Madison wrote whilst; Hamilton wrote while. Madison reached for upon far less often. Tiny unconscious habits — enough to assign all twelve papers with confidence that still holds.

A voiceprint is the Federalist method run backwards: measure a person, then project that fingerprint forward onto a model.

Writer A Writer B the of upon while whilst by
Common words match almost exactly. The tell is in the throwaways no one chooses on purpose.
Biber's dimensions & the Big Five

Style is a set of dials, not a set of boxes.

In the late eighties the linguist Douglas Biber ran a mountain of text through factor analysis and found that features cluster into continuous dimensions. The sturdiest runs from involved (contractions, present tense, "you") to informational (dense noun phrases, the writer stepping out of frame). A model already knows what every one of these means — so you can hand it your coordinates.

Personality leaves marks too. Lexical diversity, the ratio of adverbs to adjectives, how much you hedge — the Big Five traits show up as measurable habits. Timbrel maps all of it across thirteen dimensions.

InvolvedInformational
NarrativeAnalytic
ExplicitSituational
Rarely hedgesHedges often
You — a distinct spot The model — always mid
The model crowds the safe center of every dial. You don't — and that's the coordinate we capture.
The anti-AI filter

The tells a voiceprint hunts down.

A model has a fingerprint of its own — habits it can't stop performing. Every voice file bans them by name, on by default, even in the free tier. These are the usual suspects.

delve · tapestry · realm
Words that barely exist in human prose. The model mostly learned them from other models.
the rule of three
Bundling everything into a tidy trio whether or not the third item has anywhere to be.
Moreover, Furthermore
Connective scaffolding bolted on when the logic was already obvious without it.
not just X — it's Y
A fake-profound inversion that borrows the cadence of insight and pays out nothing.
a neat bow per paragraph
Paragraphs swelling to identical size and tying off on identical little summary lines.
the em-dash reflex
Parenthetical asides bolted on by habit — banned unless the em-dash is genuinely yours.
And your own genuine quirks? The one phrase you honestly overuse is marked exempt, so it survives the cull.
The output

All of it, folded into one file.

A voiceprint is plain Markdown you paste into any model's system prompt. The order is deliberate: models pay most attention to the start and end of a prompt and lose the middle — so the hard rules are bracketed top and bottom. That same repetition is what keeps the voice from drifting as a long chat fills with noise.

The few-shot examples are the sample passages you picked as most like you — the model learns your voice by demonstration, never from prose you had to write. Ready-made context variants — email, blog, social — ride along on Pro.

⛔ Hard rules · pinned to the top
Write as <you> — three exact adjectives. The non-negotiables, first.
bracketed top & bottom — beats "lost in the middle" & drift
Identity
Who the voice is, and the exact lens it writes through.
<forbidden>
The tells kill-list — with your signature phrases marked exempt.
Style vectors
Biber + Big Five coordinates, written as plain instructions — never jargon.
<example>
Two passages you picked as sounding most like you. Voice by demonstration.
re-anchor
⛔ Hard rules · repeated at the bottom
The same non-negotiables, so the last thing read is the voice.

Measured, not guessed.

That's the whole idea. Spend about twelve minutes mapping your own coordinates and hand your AI something no one else can wear.

Build your voiceprint
A voice rendered as a soft, undulating soundwave ribbon — a finished voiceprint, measured not guessed