Tone of voice generator

By Jack Stovell X · Instagram · LinkedIn · published 2026-08-03 · updated 2026-09-30

Stop writing "friendly but professional" and hoping for the best

ScriptGrain measures the tone you already have and turns it into something your writers can't misread.

Generate a tone of voice prompt from your own writing, free. Paste 120 words or more that you wrote yourself. The builder below measures them the way ScriptGrain measures a profile and writes a one-page tone of voice prompt in numbers: sentence length, contractions, punctuation, plus formality and humor as a judge reads them. The prompt is yours to copy. ScriptGrain itself connects to Claude, ChatGPT, Cursor, Meta Muse and Gemini CLI through its MCP server, which reads your full measured profile; the Gemini app takes this prompt instead. Nothing you paste is stored.

Some tools take a different route to tone. Anyword detects it from sample text or a page, then pairs it with brand rules and approved phrasing. We measure 45 style attributes from your writing instead, show every number, and score each draft against the profile feature by feature. The ScriptGrain vs Anyword comparison lays out where each approach fits.

The problem with adjective TOV documents

"Friendly but professional" means nothing. Give that phrase to three writers and you'll get three different voices back, and none of them will sound like you. Adjectives feel useful. They're not enforceable. That's the whole issue.

Most tone of voice documents are a page of vibes. Nobody can action a vibe.

Measurement is the fix

So instead of guessing at adjectives, ScriptGrain measures your actual writing. Two-pass stylometric analysis, 45 attributes pulled from real samples: lexical choices, sentence structure, tone markers, rhetorical habits, punctuation, function words, content patterns, the odd quirks that make your brand sound like your brand and not a competitor's. Data, not vibes.

How it works, three steps

  1. Upload samples. Drop in existing copy, five pieces is plenty.
  2. Get your profile. See all 45 attributes measured and explained, no guessing what "professional" means to your team.
  3. Write in it. Generate content locked to the profile, with every draft scored against your measured baseline.

What's actually in the profile

Not adjectives. Constraints. Your measured contraction rate, sentence length variance, comma habits, preferred transitions, the specific words you reach for and the ones you never touch. Hand it to a freelancer, or generate straight from it. Either way it's specifics, not opinion.

That's why two writers briefed from the same profile land in the same voice, not two different ones.

Free to see it, paid to use it

Your first profile and full analysis: free, no card required. See exactly what your tone measures across all 45 attributes before you pay anyone anything.

Want to generate content in that voice? Your first draft is free; writing in it regularly is on the paid plans.

Fair enough, you might say, but why pay at all if the profile's free? Because measuring your tone is one job. Generating on-brand copy at scale, across a team, without drift, is a different job entirely. That's the one we charge for.

Get your tone measured

No card. One profile, one full analysis, yours to keep.

Still explaining "friendly but professional" to a new hire? Stop. Measure it once, write in it, done.

What a tone of voice guide should contain

A useful guide does three jobs. It shows the voice, it fences the voice, and it gives the voice numbers. Most documents manage the first, skip the second and never get near the third.

Start with examples. Real ones, lifted from your own published copy. A writer learns more from one genuine paragraph than from a page of description, because they can copy the rhythm without having to decode it. Add a short "this, not that" pair so the edges are visible: "We moved the launch to April" sits inside the voice; "We are delighted to announce a revised launch window" doesn't.

Then the dos and don'ts. Keep them concrete. "Do open with the conclusion" is a rule someone can follow. "Do be engaging" is a wish. The same goes for bans: name the words you never touch and the openers you'd cut on sight.

And then the habits, in numbers. Average sentence length and how much it swings. How often you use contractions. Whether you reach for semicolons or avoid them. How formal you are on a scale of 0 to 10. These are the things a reader feels without being able to name, and they're the things a writer gets wrong first.

So, a guide worth keeping has examples, rules and measurements. Drop any one and the other two wobble.

From tone words to measurements

Take "warm but authoritative": every writer believes they're doing it. Hand the same phrase to three of them and you'll get three drafts, each sincerely on brief, each different. The adjectives don't fail because writers are careless. They fail because an adjective has no edges.

A measurement does. Say your average sentence runs at about 17 words, but with big swings, so a four-word fragment sits next to a thirty-word build. That's a shape a writer can hit or miss. Say you use contractions freely, and you open paragraphs with the point rather than the context. Again: checkable.

ScriptGrain's profile covers 45 attributes across 8 layers: lexical, syntactic, tone and register, rhetorical, punctuation and format, function words, content patterns, and quirks and cadence. Sentence length and variance, contraction rate, comma density, semicolon and dash habits, confidence versus hedging, argument structure (evidence-first, conclusion-first, narrative), pronoun mix. Each one is a number or a named pattern, not a mood.

It also works as a test afterward. Every draft gets a voice match score from 0 to 1. On-voice drafts typically land between 0.85 and 0.95; partial matches sit between 0.5 and 0.8; a clearly different voice falls under 0.45. The score comes with per-feature deltas, draft value against profile value, so "it feels off" becomes "your sentences run longer than your baseline and you've hedged twice as often."

That's the difference. Feedback you can act on.

Using the guide with writers and assistants

For a human writer, keep the handover short. Send the examples first, the measured habits second, the rules last. Ask for a sample paragraph before the full piece, then compare it to the numbers rather than to your gut. A freelancer who misses on sentence length can fix sentence length. One told to "sound more like us" can only guess again.

For an assistant, the method matters more than you'd expect. The prompt this page builds is a good start, but a pasted block is a snapshot: it sits at the top of a conversation and can get buried by the fourth reply. The full profile works differently. It reaches ChatGPT, Claude, Gemini, Cursor, Meta Muse and any MCP client through the MCP server and the REST API, and it's read live each time. Both are on every plan, including Free. The assistant is checking the profile, not recalling a prompt.

A few habits make either route work better:

  1. Give three or more pieces and around 3,000 words when you build the profile; one sample of 50+ words will start it, but more gives it more to measure.
  2. Use writing you actually published, not writing you wish you'd published.
  3. Re-run the score on every draft and read the deltas before you read the prose.

If a draft lands short of target, Polish revises it toward a score you set (0.9 by default). And the app learns from the edits you save, so the guide sharpens as you use it.