Consent first: the ethics of using AI voices

AI voices make it possible for anyone to produce professional narration in minutes. The same technology could also be used to put words in someone else's mouth. How a platform and its users handle that difference decides whether AI voice is a tool people trust. This article explains the principles we follow at Vocla and the practical rules they lead to.

A voice belongs to a person

A human voice is part of someone's identity. People recognise friends, family and public figures by voice alone, and they tend to believe what they hear. That is why we treat a real person's voice as theirs, in the same way you would not use someone's photo in an advert without asking. The principle applies whether the voice is famous or not: a colleague, a voice actor and a head of state deserve the same protection.

How Vocla approaches voice cloning

Voice cloning, meaning recreating a specific real person's voice from recordings, is the feature most open to abuse: a convincing copy of a real voice can be misused once it exists. That is why Vocla offers it on paid plans only under clear rules:

• Consent first. You may clone only your own voice, or the voice of someone who has given you explicit written permission.
• Private by default. A cloned voice stays private to the workspace that created it and is never listed in the public voice library.
• Delete anytime. You can delete a cloned voice whenever you want, and deleting it also deletes the underlying voice model.
• No impersonation. Cloning public figures, celebrities, officials or anyone else without their consent is prohibited. We act on reports, and we may remove cloned voices and suspend accounts that break the rules.

The process itself is simple: you upload or record short clips of clean speech (from about 10 seconds up to 5 minutes in total), Vocla builds a private voice model, usually in about a minute, and that voice can then read any text in text to speech.

Voice design: new voices, not copies

When you do not need a specific person's voice, Vocla offers voice design: you describe the voice you need, for example "a warm, middle-aged Saudi narrator with a calm pace", and Vocla proposes several original synthetic voices that match the description. You can also point to library voices as style references. Those references guide the tone and delivery; they do not copy anyone's identity.

For most real projects, an original voice is what you want anyway. It is consistent, it belongs to your project, and nobody else can claim it is theirs.

Impersonation is off limits

Some uses are not acceptable regardless of how the audio is made:

• Making a real person appear to say something they did not say.
• Imitating public figures, officials or celebrities, including parody that could be mistaken for the real thing.
• Fraud and social engineering, such as fake calls from a bank, an employer or a relative.
• Content that harasses, threatens or deceives.

Our acceptable use policy prohibits these uses, and we check text against blocked patterns before generation. No automated check is perfect, so we also act on reports.

Consent still matters when real people are involved

Real people can appear in your audio projects in other ways too: an interviewee whose recording you clean up, a presenter whose delivery you convert with the voice changer, a colleague who reads the intro. Ask before you publish their voice, tell them how it will be used, and keep their permission on record. Our voice consent policy explains the rules for audio you upload.

Tell your audience when it matters

Not every use of an AI voice needs a label. A phone menu or an app notification is obviously automated. But listeners deserve to know when the distinction matters to them:

• Narration presented as a person's own reading, such as an author narrating their book.
• News, testimonials and interviews, where listeners assume a real human is speaking.
• Advertising that features a "customer" or an "expert".

A short note such as "narrated with an AI voice" in the description is usually enough. Some platforms and regulators have their own disclosure rules; check the ones that apply to you.

Library voices and your rights

Vocla's public library voices are provided for you to use. On the Professional and Enterprise plans, audio you generate can be used commercially under our terms. What you cannot do is present a library voice as a specific real person, or use it for anything our acceptable use policy prohibits.

How Vocla keeps generated audio accountable

• Generation records: we keep a fingerprint registry of audio generated on Vocla, so we can help answer the question "was this made here?".
• Moderation before generation: text is checked against blocked patterns before credits are spent.
• Reporting: anyone can report misuse through the contact page by choosing the privacy topic.

A checklist for responsible use

1. Use library voices, voices you designed, or voices you have the right to clone; never recreate a real person's voice without their consent.
2. Never make a real person appear to say something they did not say.
3. Get permission before publishing recordings of real people.
4. Disclose AI narration where listeners would assume a human.
5. Report misuse when you see it.

Responsible use is not a limitation on creativity. It is what keeps listeners willing to trust what they hear, and that trust is what makes voice content valuable in the first place.