How to Create a Custom AI Voice in OndaLabs

To create a custom AI voice in OndaLabs, you need the Creator plan or higher. Then you can either clone your own voice from a short recording and a live consent sentence, or design a new voice from a written description. Each option costs 1,500 credits and uses one of your plan's voice slots.

This guide covers both methods, the consent steps, the plan limits and some tips for better results. Field names are quoted as the app shows them.

Clone or design: which should you create?

Both methods make a saved voice that you can use in text to speech and dialogue. They start from different inputs.

| | Voice cloning | Voice design | |---|---|---| | Starting point | Your recording, 10 to 30 seconds | A written description, 10 to 1,000 characters | | Extra input | A consent recording of an on-screen sentence, plus a tick box | An optional gender: Any gender, Female, Male or Neutral | | Cost | 1,500 credits | 1,500 credits | | Plan | Creator and up | Creator and up | | Time | Not stated in the app | About 30 seconds, per the app | | Result | A "cloned" voice in My voices | A "designed" voice, with an audio sample when it is ready | | Best for | A familiar voice, such as your own narration | A new character or brand voice that does not exist yet |

What you need before you start

A plan with custom voices. Creator allows 3 voices, Pro allows 10 and Business allows 30. Free and Starter show a lock panel with an Upgrade button.
1,500 credits per voice. On Creator's 400,000 monthly credits, one voice is 0.375 percent of the month (1,500 divided by 400,000).
For cloning: a quiet room, one speaker, a microphone you can use for both recordings, and 10 to 30 seconds of natural reading.
For design: a description of at least 10 characters and no more than 1,000.
Microphone permission. If you deny it, the app shows "Microphone access was denied."

How to clone your voice

Open My voices and choose the Voice cloning card. The subtitle reads "Two recordings in a quiet room, with the same microphone."

Enter a voice name of up to 50 characters.
Record 10 to 30 seconds of yourself reading anything naturally, or upload an audio file. Recording stops automatically at 30 seconds. Uploads can be formats such as .webm or .m4a.
Pick the language, then read the on-screen consent sentence for that language. Record it as the consent recording, up to 30 seconds, or upload it.
Tick the box that reads "This is my own voice, I am 18 or older, and I consent to creating a synthetic model of it."
Submit. The button stays disabled until the name, both recordings and the checkbox are all complete. On success, the app shows "Your voice is ready."

The clone dialog is built for your own voice. Some marketing copy talks about recordings you have permission to use, but the steps above ask you to confirm the voice is yours.

How to design a voice from a description

Choose the Voice design card. Enter a voice name of up to 50 characters, then describe the voice. The description needs at least 10 characters and allows up to 1,000. The product's placeholder shows the level of detail it expects:

"e.g. A warm, thoughtful astronomer in his late 60s with a gentle British accent, speaking with quiet wonder."

Choose a gender if it matters: Any gender, Female, Male or Neutral. While the voice is being designed, the button reads "Designing (about 30 seconds)...". When it finishes, you hear a sample and can press Done to save the voice.

A description that works well usually covers five things, in this order:

Age
Accent
Tone
Personality
Pace or setting

This order is our suggestion, based on the app's own hint, "age, accent, tone, personality."

Consent and responsible use

Cloning is the part that needs the most care. The clone dialog asks you to confirm three things: the voice is yours, you are 18 or older, and you consent to a synthetic model of it. It also asks you to record the consent sentence live.

The terms say you may not create voices that impersonate real people without their consent, and you may not upload copyrighted material without the rights to it. OndaLabs also recommends being open about AI use when you share content.

Consent checks vary across the industry. Consumer Reports' 2025 assessment of six voice cloning products found that four of them had not taken basic steps to stop unauthorized cloning. It recommends "adding a technical mechanism to confirm a speaker's consent to having their voice cloned."

Proof News reported in June 2024 that some tools asked users to read a script that was checked by transcription, that a tester bypassed one such check with another cloning tool, and that most tools relied on a checkbox. A live consent sentence is a useful check, but it is not a guarantee. This guide does not claim that OndaLabs's check cannot be bypassed.

The Partnership on AI's Responsible Practices for Synthetic Media asks creators to be transparent about how consent was obtained and to disclose synthetic elements, either with a label or with provenance information. Labelling your output is a good default.

Use your custom voice

A saved voice appears under My voices in the voice picker of text to speech, and as a speaker voice in dialogue. Select Use on a saved voice to choose it and open text to speech. The dialogue page notes that cloned and designed voices work there too.

Commercial use is included from Starter, and Creator includes everything in Starter. Custom voices start at Creator, so commercial use is available for the voices you create on that plan or higher.

Plan limits and managing saved voices

Cloned and designed voices count toward the same slot limit.

| Plan | Price per month | Custom voice slots | Credits per month | |---|---|---|---| | Free | $0 | 0 | 2,000 | | Starter | $9 | 0 | 170,000 | | Creator | $22 | 3 | 400,000 | | Pro | $49 | 10 | 900,000 | | Business | $199 | 30 | 3,750,000 |

The My voices page shows a counter, "n / limit". When you reach the limit, both creation cards are disabled, and the page says "You've reached your plan's voice limit. Delete a voice or upgrade."

To free a slot, delete a voice. The app asks you to confirm with a message like "Delete" followed by the voice name and "This can't be undone." Once deleted, the voice is removed from the list and its slot is free for a new one.

Tips for better results

Record in a quiet room with no echo, and make sure only one person is speaking.
Use the same microphone for both the voice recording and the consent recording.
Read at a steady, natural pace for the whole 10 to 30 seconds. Do not rush it or perform it.
Be specific when you design. Lead with age, accent, tone and personality, then add pace or setting.
Test before you commit a project. Generate a paragraph with the new voice, listen to it, and only then use it in a longer script.

A short sample is not a promise of a particular quality. Listen to the result, not the description.

Where to go next

If you are choosing between methods for a project, the pricing guide shows how the 1,500 credit cost fits into a month. The model comparison helps you choose the model for the voice's first full script. You can try a saved voice in text to speech or in dialogue.

Frequently asked questions

How do I create a custom AI voice?

Open My voices on a Creator plan or higher, then choose cloning or design. Cloning needs a recording of 10 to 30 seconds, a consent recording and a checkbox. Design needs a description of 10 to 1,000 characters. Both cost 1,500 credits.

What is the difference between voice cloning and voice design?

Cloning copies a voice you record yourself. Design creates a new voice from a written description, and the result comes with an audio sample. Use cloning for a familiar voice, such as your own narration, and design for a new character or brand voice.

How long should a voice sample be for cloning?

The clone dialog asks for 10 to 30 seconds of natural reading, and it stops recording automatically at 30 seconds. The app does not state a quality level for a given length, so this guide does not promise one. Clean audio from one speaker, recorded consistently, matters more than length alone.

Do I need consent to clone a voice?

Yes. OndaLabs requires you to confirm that the voice is your own, that you are 18 or older, and that you consent to a synthetic model of it. The terms also prohibit creating voices that impersonate real people without their consent. For legal questions, check the law where you live and where you publish.

How many custom voices can I have?

Creator allows 3, Pro allows 10 and Business allows 30. Free and Starter allow none. Cloned and designed voices share the same limit, and deleting a voice frees its slot.