Sometimes you need a voice that does not belong to anyone in particular — a character, a narrator, or a brand tone you can reuse. The free AI voice designer on ToolXoX builds that voice from a written description. You do not upload a recording. You describe the sound, set a preview line, and click Generate Voice. There is no account to create.
A designed voice is different from a cloned one. Cloning copies a real person from a sample. Designing invents a new synthetic voice from your prompt, which is the safer path when you do not have a recording or permission to imitate someone.
What you write in the prompt
The description box holds up to 1,000 characters. The useful details are age, gender, energy, and emotion, plus the job the voice should do — narration, a story, an ad, a calm guide. The page also offers ready-made examples you can start from: Energetic Storytelling, Confident Narration, Calm & Soothing, and Warm Casual. Click one to fill the box, then edit it so it matches your project.
A specific prompt beats a vague one. “A calm young woman, speaking slowly, gentle tone, for a relaxation guide” is easier to judge than “a nice voice.” You can write your own description from scratch if none of the examples fit.
How to design a voice, step by step
- Describe the voice in 1,000 characters or fewer, or start from one of the example cards and adjust the wording.
- Open Text to Preview. Choose the language and, if you want, edit the sample sentence (up to 300 characters). That is the line you will hear when the voice is ready.
- Complete the security check and click Generate Voice once. The queue status is “Creating your custom voice.” Leave the tab open until the audio is ready.
- Listen to the preview and download it if the tone, pace, and language match what you asked for. If they do not, change one part of the description — energy, age, or use case — and generate again.
How to write a description that holds
Name the speaker, the feeling, and the setting. A deep, steady voice for a documentary and a bright, quick voice for a short promo should not share the same prompt. Mention the language you will actually use in the preview, because a beautiful English take will not help if the final script is in another language.
Do not try to describe a famous person, a specific actor, or “someone who sounds exactly like” a real speaker. That turns a designed voice back into an imitation problem. Stay with traits and roles, not identities.
When to design, and when to use something else
Design a voice for characters, brand narration, and any project where you want a consistent tone without copying a person. If you already have a clean recording of yourself and you want that exact sound, cloning is the closer match. If you only need a one-off read of a script and a stock voice is enough, text to speech with a ready-made voice is usually faster.
For audio that is not speech, use the AI sound effect generator on CanvaPro, or follow the steps to generate a sound effect from text. If you are cleaning a finished mix rather than inventing a voice, the vocal remover splits singing from a backing track, and the vocal remover guide walks through that job.
The four examples are a useful starting grid. Energetic Storytelling is a bright, expressive young female voice for audiobooks. Confident Narration is a deeper, formal male voice for documentaries. Calm & Soothing is a slow, gentle voice for relaxation. Warm Casual is a friendly, upbeat male voice for everyday or promotional lines. Start there, then change the age, the pace, or the use case rather than rewriting the whole prompt from nothing.
If the voice misses the brief
An empty description, a preview with no text, or an expired security check will stop the job. Fill both boxes and try again. If the voice is close but too bright, lower the energy in the prompt and keep everything else the same. If the language of the preview is wrong, change it in Text to Preview rather than stuffing a language name into a prompt that is already long.
Save the prompt with the download. A voice you like is hard to reconstruct later from a vague memory of “the calm one.”
Keep the preview sentence ordinary and in the language of the final script. A line full of invented names is harder to judge than a short paragraph the voice would actually say on your project.
When you know the tone you want, open the custom voice designer, start from an example or your own description, and download the preview only after it sounds like a voice you would actually put under a script.