How to use this guide
This guide helps you understand how you can use our voice technology safely, follow best practice, and respect our terms of service. What constitutes a lawful use of voice cloning may vary depending on who created the content you are reusing, where you are located and where you publish your materials. This guide describes what constitutes safe uses of voice cloning in most countries, with a focus on countries in the European Union and the United Kingdom.
This guide does not constitute legal advice.
What do we mean by voice cloning?
“Voice cloning” is the digital imitation of a person’s voice with high fidelity results. When you use our voice cloning technology you upload a recorded voice, and our model creates its digital representation. You can then use this digital
representation to generate new speech.
Our cloning technology produces voices which sound similar but are not technically identical to the materials you have uploaded.
Voice cloning: safe use and best practice explained.
You can voice clone in three scenarios, with exceptions described below:
- Voice cloning will be safe if you use your own voice and content to generate a clone. You can use this clone for any purpose.
- Voice cloning will be safe if you use someone else’s voice or content to generate a clone with their permission. You can use this clone for any purpose you received permission for.
- Voice cloning is only sometimes safe if you use someone else’s voice and materials to generate a clone without their permission. You can use this clone for certain non- commercial purposes if you don’t impact the person’s privacy or economic interests.
These non-commercial purposes include:
- private study and non-commercial research
- education
- caricature, parody and satire
- artistic and political speech contributing to public debates.
- quotation
- criticism and review
A very important rule applies to all uses of voice cloning technology: you cannot clone a voice for abusive purposes such as fraud, discrimination, hate speech or for any form of online abuse without infringing the law.
This rule applies even if you use your own materials to generate a clone or have received others’ permissions. If you use voice cloning technology for harmful purposes, you may be criminally liable. As a service provider, we may be obligated by law to report criminal and other unlawful activities occurring on our sites. We may also suspend or terminate your
account.
Cloning a voice safely, and following best practice, is a simple two-step process you can follow yourself. We describe what this process looks like in practice in the next section.
Voice cloning: safe use and best practice in two steps
There are usually two steps involved in working with voice cloning technology: making the voice clone and sharing it. We explain each step below and give you tips on what you can do to make things easier.
Step 1 | Make your voice clone.
You need three key source materials to generate your voice clone:
- a voice
- a recording of that voice
- input text to generate new speech.
Each source material may carry different rights. Typically, a voice is protected by rights such as personal data, personality, publicity or performers’ rights. The recording may carry copyright in the spoken words and in the recording itself. The “input” text can also be protected by copyright.
If you are using your own voice and have created your own recording and text, you can safely use them to generate your voice clone. This is because as the creator you will own the rights in your materials. If you are using other people’s content, you will need to “clear” the rights in the materials to be able to use them for voice cloning. “Clearing rights” means asking for permission to use the materials from the people who own them or checking that the law allows you to use the materials without permission.
Step 2 | Share your voice clone.
Credit your source materials
You should always credit the creators of the source materials you have used to generate your voice clone (including yourself!). Be as precise as possible in your credits. Indicate whose voice you have cloned, who produced and edited the sound recordings, and who wrote the text you’ve used to prompt new speech in the voice clone.
Disclose content manipulation.
Make it clear that your voice clone is manipulated content, and as such it does not constitute an “authentic” recording of the imitated voice. You can do so by adding a notice in the text or images accompanying the voice clone when you share it - for example by a simple mention that your content is AI-generated. You can also make your voice clone read your notice at the start or the end of the recording by including this information in your “input” text.
Communicate your own rights and permissions in the voice clone.
Your voice clone may be original enough to be protected by its own intellectual property rights. If this is the case, you may be the owner of the rights in the voice clone. You may also own the rights in other source materials you have created, like the text you have used to prompt new speech.
It is always useful to tell others what they can or cannot do with your voice clone when you share it. If you are happy for the public to use your voice clone, tell them! Explain whether you have conditions for the use of your voice clone by others. For example, are you happy for people to use your voice clone for commercial and non-commercial
purposes? Can other people modify and repurpose your voice clone?
Do you want to be credited as the creator of the voice clone? If you clone your own voice and share it under a free license, your license will also allow others to re-clone your voice to generate new speech. If you do not want this to happen, make sure you exclude this permission from the scope of your license. While you are the owner of your materials, you also give us a license to your contributions (including your voice clone) when you upload or connect them to our sites. For more on this,
revisit our terms of service.