Stop re-reading old client files just to remember how they sound
Measure the voice properly. Once. Then let it hold.
Here's the bind you already know
"Professional but warm" describes every client you've ever had. So do the three adjectives in their brand guidelines. None of it tells you whether they use semicolons, how long their sentences run, or whether they'd ever say "leverage" out loud.
You know this already; that's why you re-read six old drafts before starting a seventh. That's the job nobody bills for.
What ScriptGrain actually does
It measures 45 stylometric attributes from writing your client has already published: sentence rhythm, contraction rate, punctuation habits, formality, humour register, hedging versus declarative claims, the lot. Paste it, upload it, or pull it straight from their website with brand mimic (with their consent, obviously).
One profile per client. Not a mood board. A measurement.
Drafts that hold the line
Every draft gets built against that profile, across 16 content types: long-form, newsletters, LinkedIn, email, ad copy, press releases, landing pages, more. Each type has its own structure spec. The voice constrains all of it.
A score before anything ships
You get a voice-match score with per-feature reasons before the client ever sees the draft. Not "does this feel right?" Something you can actually check.
A polish loop, not a rewrite from scratch
If a draft falls short, the polish loop revises it toward the measured profile using the actual divergences, not vague notes to self. You stay the editor. You stay the expert. The tool just stops guessing.
It learns from what you keep
Accepted edits feed closed-loop learning. So the profile's drafts get closer to the client over time, based on what you actually changed, not what a generic model assumes "warm" means this week.
Why this matters more with AI in the room
Here's the thing: AI tools are why this problem got worse, not better. Every client starts sounding the same beige mid-Atlantic voice, and clients notice. They just can't always say why.
ScriptGrain doesn't replace your judgement. It gives the machine something specific to aim at, so what comes back is closer to right before you touch it. You're still the one who knows when a line is off. You just spend less time getting back into character first.
Built for people juggling more than one voice
Ten clients, ten profiles, ten distinct voices measured properly instead of ten sets of adjectives you're trying to hold in your head at once. British English throughout, if that's where your clients live.
Try it without the faff
Free tier: one voice profile, one full analysis, one free draft. No card required. See what "measured" actually looks like against a client you already know.
Paid plans start at £12/month for a single profile with 40 generations. £29/month gets you three profiles and 75 generations, useful the moment you're juggling more than one client at once. The 10-profile plan runs £99/month, for the writers with a full roster and no patience left for re-reading old drafts.
Fair enough if you're sceptical. Most writers are, about anything claiming to understand voice. Measure one client's writing and see if it disagrees with you.
Questions
How do I build a profile for a client?
Paste or upload writing the client has already published, or import it from a URL. Brand mimic can build the profile straight from their public website, with their consent. Either way it's one profile per client, measured across the same 45 attributes.
How many client voices can I run at once?
One profile on the £12/month plan, three on £29/month, ten on £99/month. Each is a separate measured voice, so switching clients means switching profiles, not re-reading old drafts to get back into character.
How do I know a draft matches the client before I send it?
Every draft returns a voice-match score with per-feature reasons: which attributes held and which drifted. If it falls short, the polish loop revises it toward the measured profile and rescores, so what reaches the client is checked, not hoped.
Do the drafts improve as I edit them?
Yes. Accepted edits feed closed-loop learning, so the profile's future drafts move toward what you actually changed rather than what a generic model assumes the brief meant.