How do I get a genuinely original brand voice, not a real person's?

This page may contain affiliate links from which Blur Studio earns commissions.

A real human voice comes with real human complications: rights, cost, availability, and the fact that the person behind it exists and has opinions about how their voice gets used. A voice that was never a real person to begin with sidesteps all of that, which is exactly why a brand mascot or signature AI voice is often designed from scratch rather than cloned or licensed from an actual person.

How voice design actually works

ElevenLabs' Voice Design builds a new voice from a written text description rather than a recording, and it's available even on the free plan, so there's no cost barrier to trying it. Each time you generate from a prompt, you get a handful of preview options to choose from, not just one, so you're comparing a few interpretations of your description rather than accepting the first result. A setting that controls how closely the output follows your exact wording versus taking creative liberty is also worth knowing about: tightening it keeps results closer to your description, loosening it gives more variety at the risk of drifting from what you asked for.

What to actually write in the prompt

Demographics alone, "a 40-year-old man," produce a generic result. Personality and delivery get you much closer to something distinct: warm but efficient, playful without being silly, calm and unhurried, a voice that sounds like it's smiling. Describe how the voice should make a listener feel, not just what it technically is. If you have a real reference in mind, a tone from a show, a podcast host's energy, describing that feeling in your own words works better than trying to name the reference directly.

Step by step

1. Write a short description focused on personality and delivery, not just age and gender.
2. Generate and listen to the preview options from that one prompt before changing anything.
3. Adjust the wording, not just the settings, if none of the previews feel right, since the description itself usually needs sharpening before a setting tweak will help.
4. Once you have a voice you like, test it on a real line from your actual brand content, not just the preview text, before committing to it everywhere.
5. Save the final settings and prompt so you (or whoever edits content later) can regenerate a consistent voice if you ever need to.

A before and after prompt, worked through

A vague prompt like "friendly female voice, 30s" produces something generic enough that it could belong to almost any brand, because it's describing a category of voice rather than a specific one. A sharper version of the same idea: "a warm, unhurried voice that sounds genuinely glad to hear from you, like a friend who happens to work in customer support, confident but never rushed, with a slight smile audible even when explaining something technical." That second prompt gives the generation something concrete to work from: pacing cues ("unhurried," "never rushed"), emotional tone ("genuinely glad," "slight smile"), and a relatable reference point ("a friend who happens to work in customer support") that a generic demographic description can't capture. Notice the second prompt never mentions age or gender directly, since personality and delivery almost always matter more to how a voice actually lands than the demographic details people default to describing first.

If you're designing a voice for a specific function, a customer-support agent versus a sales demo versus an audiobook narrator, write the prompt around that function specifically rather than a generic "nice voice" brief. "A patient, reassuring voice for someone frustrated with a technical problem" produces a meaningfully different result than "an energetic, upbeat voice for a product announcement," even if both could technically be described as "friendly." For a cast of several voices, see making different character voices for a course or game.

Keeping the voice consistent once more than one person is generating content with it

Once a designed voice becomes your brand's standard, more than one person on your team is probably going to be the one actually generating audio with it eventually, and that's where consistency problems creep in if you haven't planned for it. Save the exact voice (not just the prompt that created it) as the single source of truth, and document the generation settings that worked for your typical content, rather than letting each person who touches it experiment with the knobs independently. A voice that sounds warm and controlled in one piece of content and slightly manic in another, because two different people generated them with different settings, undercuts the exact consistency a designed brand voice is supposed to provide over a library voice anyone could pick. More on this is in keeping the same AI voice in every company video.

The one setting most people never touch

Beyond the description itself, there's usually a control for how strictly the generation follows your exact wording versus taking creative liberty with it. Set low, you get more variety across attempts, which is useful early on when you're still exploring what the voice should sound like. Set high, results stick closer to your description but can start sounding slightly artificial or forced if your description is short or vague. The practical rule: write a longer, more detailed description and use a lower setting for genuine exploration, then tighten the setting once you've found a direction you like and want to nail down consistently. Most people never touch this control and just regenerate the same short prompt repeatedly, which is a slower way to get to the same result.

Can you actually own a designed voice, the way you own a logo?

Not in the same way, and it's worth setting expectations honestly here. A logo can be registered as a trademark because it's a specific, fixed visual mark. A voice, even a designed one that was never a real person, isn't automatically protected the same way just because you generated it and started using it in your branding. Someone else describing a similar personality could land on a similar-sounding result, as covered above, and there's no registry you file a designed voice into the way you'd file a trademark application for a logo or brand name. What you can reasonably protect is the specific combination of the voice plus how you consistently use it, your actual recorded content, your brand guidelines describing how the voice should sound and behave, which functions more like a style guide than exclusive ownership of the raw voice itself. If having a legally exclusive, unmistakably-yours voice matters enormously to your brand, the more defensible path is a custom-recorded voice actor under an exclusive agreement, not a generated one, even though that costs considerably more.

Designed voice versus a library voice versus a real clone, for brand use

These three options solve slightly different problems, and it's worth being clear on which one you actually need. A stock library voice is the fastest and cheapest: pick one, use it, done, but you're sharing it with however many other brands and creators also picked that exact voice, so it carries no real distinctiveness. A designed voice, the subject of this article, costs a little more effort upfront (writing and refining a prompt) but gives you something built specifically around your brand's personality, even though it's not exclusively yours in a legal sense. A real human voice clone, your founder's or a hired voice actor's, gives you the most genuine distinctiveness and an actual person's rights to manage, but comes with the real-person complications (consent, departure risk, licensing terms) this article opened by pointing out. Most brands that don't have a strong reason to clone a specific real person land on a designed voice as the right middle ground: more distinct than a stock pick, without the ongoing dependency a real person's voice carries.

What I'd tell you before you lock this in

Don't assume a designed voice is exclusively yours. Other users describing something similar can land on a comparably sounding voice, since the tool is generating from a description, not reserving a unique output for you. For most brand use, this is a non-issue; for a flagship mascot voice you want to feel truly proprietary, consider having the designed voice professionally recorded or significantly customized afterward, so it's no longer just the raw generated output anyone else could land near.

Preview text in the tool is often short and limited, which means a voice that sounds great on the preview can behave differently on your actual, longer script. Always test your real content before deciding, the same rule that applies to every voice choice on this platform, designed or not. For a fast way to try lines out, see making quick voice drafts for a pitch demo.

When to skip this and build something bigger

If the brand voice needs to hold real conversations, not just read pre-written lines, a designed voice is only the starting point, not the whole product. If that's actually where you're headed, tell me what the voice needs to do and I'll tell you honestly whether a designed voice and a script are enough or you're building toward a full voice agent.