Design a Voice mode lets you create an original AI voice with nothing to clone, no recording, no real person, no consent paperwork. You describe the tone and character you want, and CoreReflex generates a brand voice you own outright. This guide explains exactly what the mode is, how it differs from cloning, how to design a voice step by step, and where that voice can work once you have it.
What "Design a voice" mode is
Design a Voice is a generative mode on CoreReflex's voice seam that builds a brand-new synthetic voice from a description rather than from a sample of someone speaking. Where voice cloning needs an existing recording to imitate, this mode starts from intent: you specify the qualities you want, gender presentation, age feel, warmth, energy, accent, pacing, and the system synthesizes a voice that matches. The result is an original voice that did not exist before and does not belong to anyone else.
That distinction is the whole point. A cloned voice is tied to a person and the permissions around them. A designed voice is a clean asset, yours to use across narration, scripts, and live conversation without depending on a voice actor's availability or a clone's consent chain. For brands that need a single, ownable sound, designing one from scratch removes a category of friction entirely.
Design a voice vs. cloning a voice
CoreReflex supports both paths, and they solve different problems. Knowing which fits your situation saves you time.
| Design a voice | Clone a voice | |
|---|---|---|
| Starting point | A written description of tone and character | An existing recording of a real person |
| Consent needed | None, no real person involved | Consent-gated; the speaker must authorize it |
| Best for | An original brand voice, a character, a fresh sound | Reproducing a specific known voice |
| Ownership | An original voice you own | Bound to the source speaker's permissions |
Cloning is the right call when you must reproduce a specific, recognizable voice, a founder, a host, an established narrator. Designing is the right call when you want a distinctive sound that is unmistakably yours and unencumbered. If you are weighing the two for your own project, the trade-offs are laid out in full in our breakdown of whether to clone a voice or design one. And if you are new to synthetic narration altogether, it helps to start with what AI voiceover is before deciding which path fits.
How to design a voice from scratch
Designing a voice is an iterative, low-stakes loop, planning and editing are free in CoreReflex, so you can refine the description and re-listen without burning through credits on every tweak. Here is the workflow.
- Define the job. Decide what this voice is for before you describe it. A voice that narrates cinematic brand films, a voice that reads punchy social scripts, and a voice that answers the phone all want different qualities. The use case anchors every other choice.
- Describe the character. Write the tone in plain language, for example, "warm, mid-30s, unhurried, lightly conversational, a touch of dry confidence." Be specific about the feeling you want, not just the demographics.
- Set the technical traits. Layer in the measurable qualities: pitch range, speaking rate, accent, and energy. These are the dials that turn a vibe into a reproducible voice.
- Generate and audition. Have the mode synthesize the voice, then listen to it reading a real line from your actual scripts, not a generic sample. The right test sentence is one you will use.
- Refine the description. Adjust the words in your brief and regenerate. "A little warmer," "slightly slower," "less formal", small wording changes move the result. Iterate until it clicks.
- Save it as a brand voice. Once it is right, store it so every pillar can call the same voice. From here it becomes a fixed asset, not a one-off generation.
Once the voice exists, you can shape each individual read with pacing and pronunciation controls. The direction layer, pauses, emphasis, and pronunciation via SSML markup, sits on top of the designed voice, so you set the character once and adjust the performance per script.
Describing tone and character: a working vocabulary
The quality of a designed voice tracks the quality of your description. Vague briefs produce generic voices. A useful description usually touches several of these dimensions:
- Warmth, cold and precise, neutral, or warm and inviting
- Energy, calm and measured versus bright and animated
- Authority, peer-to-peer and casual versus expert and assured
- Pace, deliberate and spacious versus quick and efficient
- Texture, smooth and polished versus a little gravel or breathiness
- Accent and register, regional accent, formality, and vocabulary level
Think in contrasts. "Confident but not aggressive," "friendly but not bubbly," "authoritative but not stiff" give the system clearer targets than single adjectives. The same instincts that go into choosing the right AI voice from a library apply here, except you are not picking from a shelf. You are writing the spec for one that does not exist yet.
Do you own a voice you design?
Yes. Because a designed voice is generated from your description rather than imitated from a real person, it is an original brand asset that belongs to you. There is no source speaker to license, no consent window to maintain, and no third party with a claim on the sound. That is the central advantage of designing over cloning: the voice is unencumbered from the moment it is created. You can use it as the consistent identity across every piece of audio your brand produces, and it stays yours as your library of content grows.
Putting your designed voice to work
A designed voice is most valuable because it travels. CoreReflex's voice seam is engine-agnostic and shared across the platform, so the same brand voice you designed can do every one of these jobs without sounding like three different people:
- Narrate films. The voice reads the narration over your graded cut, in character.
- Read scripts. It voices explainers, ads, and social shorts pulled from your copy.
- Answer the phone. The very same voice can power real-time voice agents and an AI appointment setter on Gemini Live, so a caller hears your brand the instant they connect.
That last point is where a designed voice quietly becomes infrastructure. When the voice that narrates your marketing is also the voice running your phone answering service, your brand sounds the same everywhere, on a landing page, in an ad, and on a live call at 9pm. One owned voice, every surface. The full set of voice capabilities is documented in the voice docs, and you can explore the rest of the Voice pillar to see how it connects to the broader studio.
Frequently asked questions
Can I create an AI voice without cloning a real person?
Yes. That is exactly what Design a Voice mode is for. It generates an original synthetic voice from a written description of the tone and character you want, with no recording and no real person involved. Because nothing is cloned, there is no consent requirement and no source speaker tied to the result.
How does Design a Voice mode work?
You describe the voice you want in plain language, its warmth, energy, pace, accent, and overall character, and the mode synthesizes a voice that matches. You audition it on a real line from your own script, refine the description, and regenerate until it is right. Once you are happy, you save it as a reusable brand voice that every pillar can call.
Do I own a voice I design?
Yes. A designed voice is an original asset generated from your specification rather than imitated from a real person, so it belongs to you with no source speaker to license and no consent chain to maintain. You can use it as your consistent brand voice across narration, scripts, and live calls.
Can a designed voice be used for live phone calls, not just narration?
Yes. The voice seam is shared across CoreReflex, so the same voice you design for narration can power real-time voice agents and appointment setters on Gemini Live. That means your brand sounds identical whether it is reading a film, voicing an ad, or answering an inbound call.
Build a voice that's unmistakably yours
Design a Voice mode turns a description into an original, ownable brand voice, no recording to clone, no consent to manage, no actor to book. Describe the character you want, audition it on your real scripts, refine the brief, and save a voice that narrates your films, reads your copy, and answers your phone in one consistent identity. Start designing yours. It is free to start, no credit card.