How many voice variants does each generation create?

Each time you click Generate Voice, Noiz AI creates three variants. Preview all three before selecting one to save.

Can I edit a voice after saving it?

You can update the voice's name, labels, and description from your Voice Library at any time. The voice audio itself is fixed at the time of generation — to change the voice characteristics, generate a new one.

What's the difference between Voice Design and Voice Clone?

Voice Clone creates a voice modeled on an existing recording — typically your own voice or a voice you have rights to. Voice Design generates an entirely new voice from a text description, with no audio input required.

Can I use a Voice Design voice in the API?

Yes. Saved voices — including ones created with Voice Design — are available in your Voice Library and can be referenced by voice ID in the Noiz AI API.

How do I find a saved voice later?

Add a clear name, labels, and a description before saving. You can then find the voice in your Voice Library and reuse it across Text-to-Speech, Studio, or the API.

How to Use Voice Design

Understand What AI Voice Design Creates

Voice Design is a generative voice creation tool in Noiz AI. Instead of selecting from the preset library, you describe the voice you want in natural language — gender, age, tone, accent, pace, personality — and the model crafts it from scratch.

Find it in the left sidebar under Playground → Voice Design, alongside Voice Library and Voice Clone.

Common use cases:

Unique voices for fictional characters in audiobooks or games
A consistent branded voiceover persona for your content
Narrator voices tailored to your audience — calm, authoritative, friendly
Prototyping different voice styles before committing to a Voice Clone

Each generation produces three variants to preview before you save one to your library.

Write Your AI Voice Design Description

Open Voice Design from the sidebar. In the text input area, describe the voice you want generated.

Be specific about tone, gender, age, accent, pacing, and personality. The more detail you give, the closer the result will match your vision. You have up to 1,000 characters to work with.

You can also upload a character image alongside your text description. The model uses visual cues from the image — age, expression, style — to inform the voice's character.

The Follow Prompt slider (0–100) controls how closely the model sticks to your description. Higher values mean tighter adherence. Start at 50 and adjust if the result feels off.

Not sure where to begin? Use the inspiration templates — Wise Elder, Sci-Fi AI, News Anchor, Dark Hero — to populate the text field with a strong starting point, then edit from there.

Pick Your Favorite AI Voice Design Variant

Click Generate Voice. Noiz AI creates three distinct voices based on your description, labeled Voice 1, Voice 2, and Voice 3.

Preview each one. When you've found the voice that fits, select it and click Select to move to the save screen.

If none of the three work, adjust your description — add more specific detail, tweak the Follow Prompt slider, or try a different template — and generate again.

Name and Save Your AI Voice Design

On the save screen, give your voice a name (required). Optionally fill in language, gender, and age range to make it easier to filter and find later.

Add a Label to tag the voice in your library, and a Description to document what it was designed for. When everything looks right, click Save.

Filling in the description field is useful if you plan to return to the voice later — it acts as a reminder of what you designed it for.

Use AI Voice Design Output in Text-to-Speech

After saving, your new voice is part of your personal Voice Library. You can select it in Text-to-Speech, Studio, or the API.

Click Use this voice to jump straight to Text-to-Speech with your new voice pre-selected, or Design another voice to start a new creation.

In Text-to-Speech, paste your script and select your saved voice from the Voice Library dropdown. It appears under My Voices.

Run a one-line test in Text-to-Speech before saving more variants — you'll hear whether the character lands in context.

Write Better Prompts for AI Voice Design

Describe a character, not a spec. Instead of "deep male voice," try "a seasoned documentary narrator in his late 40s, measured and authoritative with a slight southern warmth." Characters generate richer voices.

Mention pacing and energy. Words like "unhurried," "crisp," "conversational," or "punchy" give the model direction on delivery — not just what the voice sounds like, but how it speaks.

Specify accent and language. If you need a British RP accent versus General American versus an Australian lilt, say it explicitly. Vague descriptions tend to default to American English.

Combine image and text. Upload a character portrait alongside your description when visual cues can sharpen the result.

Generate multiple times. Slight wording tweaks between generations can yield significantly different results. Treat each attempt as a draft.

Use the Follow Prompt slider deliberately. At 50, the model balances creativity with accuracy. Push it toward 80–100 if you want it to strictly honor every detail in your description.

Try Noiz AI Voice Design ->