How to define your brand voice

You write more than you design. Emails, captions, product copy, replies, headlines, the little sentence under a button. Most of what people actually read from your brand is words, and yet voice is the part of a brand most often left undefined, described with adjectives like "friendly" and "professional" that mean nothing you can act on. This post is about fixing that: how to define a brand voice concretely enough that a person, or a tool, can follow it, and why the consistency is worth the work.

Why voice is the part that leaks

A logo appears in a few places you control. Your voice appears everywhere, often written by whoever happened to be at the keyboard, under deadline, with no reference open. That is why voice is where brands quietly fall apart. The visuals stay locked because there are files; the words drift because there is nothing to check them against.

The cost is trust. People form a sense of who you are from the accumulated texture of how you talk to them, long before they could name a single design choice. When the voice is steady, you feel like one coherent entity worth relying on. When it lurches from stiff to jokey to salesy depending on the author and the hour, you feel like a company that does not quite know itself. Consistency of voice is not a style preference. It is how a brand reads as a single, trustworthy thing.

Start from adjacent brands, not adjectives

The fastest way to make "friendly and professional" mean something is to stop using adjectives and start using comparisons. Adjectives are abstract; brands you already know are concrete.

Pick three or four brands, people, or publications whose voice you would happily borrow, and one or two you would not. Then, for each, say the specific thing you are taking or rejecting. "As direct as this one, but without the swagger." "Warm like that one, but shorter." "Never the breathless hype of that one." You will define your voice more precisely in ten minutes of comparison than in an hour of adjective-hunting, because you are pointing at real examples instead of gesturing at a mood.

A framework you can actually apply

Once you have a rough target, pin it down with four layers. Each one is meant to be written as rules, not vibes.

Character. Who is talking. Not a persona in a marketing sense, but a stable answer to "what kind of person is this brand, if it were one." Patient teacher, blunt expert, encouraging peer. This is the anchor the other three hang from.

Tone range. Voice is constant, but tone flexes with context. Define the range. How you sound in an error message is not how you sound in a launch announcement, but both should still be recognisably you. Write down the extremes so the flex stays inside bounds.

Vocabulary. The words you reach for and the words you refuse. A short list of terms you use, phrases that are yours, and, just as usefully, a banned list: the clichés, the hype words, the jargon you never want to see. Our own banned list starts with a hard no on em-dashes, which tells you these rules can be as specific as you like.

Rhythm and mechanics. Sentence length, how much you contract, whether you open with the point or build to it, how formal your punctuation runs. Mechanics are where voice is felt even when readers cannot articulate why one paragraph sounds like you and another does not.

Write all four down with examples. A voice guide with no example sentences is a wish. A voice guide with a "we say this, not that" table is a tool.

Test it against real sentences

A voice definition is only real once it can settle an argument. So test it. Take three sentences you have actually shipped and rewrite each one to match the framework, then keep both versions side by side. The before-and-after pairs are worth more than the rules themselves, because they show the framework doing work on real material rather than in the abstract.

Do this across the range, too. Write the same message as an apology, as an announcement, and as a quick reply, and check that all three still sound like the same character. If they do, your tone range is right. If one of them sounds like a different company wrote it, you have found an edge to tighten.

Keep it where the writing happens

Here is the failure that undoes all of this: the voice guide gets written, admired, and then never opened at the moment of writing. Under deadline, nobody consults a document. They just type. And a voice that lives in a document you have to remember to open is a voice that drifts.

So the definition has to travel to where the words get made. Inside Repic, your voice lives in your brand kit alongside palette and type, and it is not a reference you consult, it is context the assistant already carries. When it helps you draft, it is drafting toward your character, your tone range, your vocabulary, and your banned list by default, so the first draft already sounds like you instead of like a generic template you then have to correct. That does not remove your judgement. It removes the blank page and the drift, which are the two places voice consistency usually dies.

The point of all this

A defined brand voice is not about sounding polished. It is about sounding like one consistent, recognisable entity across every small piece of text a person encounters, so that trust compounds instead of resetting. Get it down as rules, prove the rules on real sentences, and keep them where the writing actually happens. Do that, and "sound like us" stops being an instruction nobody can follow and becomes something your whole team, and your tools, get right without being reminded.