Voice & consent policy

1. Where the voices come from

Each of Fatiman's nine voices was made from about seven to ten seconds of a speaker's recording. Every speaker signed their consent to the voice being published and used to read text aloud, including by assistants that answer customers.

Four more voices, Gason 4 to Gason 7, were made from the men's voices among the nine by changing their pitch and the shape of the voice, so that each sounds like a new person. They do not belong to any real person, and the rules below apply to them as well.

2. Good uses

• answering your customers' questions with a Lambbi assistant;
• reading your own text aloud, for yourself or your audience;
• accessibility: making written Kreyòl available to people who cannot read it easily;
• education, language learning and literacy;
• apps, announcements and media, with the audio identified as synthetic where it matters.

3. Not allowed

• presenting the voices as a real person, including the speakers themselves;
• making it seem that a real person said something they did not say;
• letting an assistant claim to be a human when a visitor sincerely asks;
• fraud, scams, voice phishing, fake emergency or official messages;
• political deception, including fake statements by candidates or officials;
• harassment, threats, defamation or hate;
• sexual content, and any content involving minors in a harmful way;
• suggesting the speakers endorse you, your product or your message.

4. Assistants say what they are

A Lambbi assistant is presented to visitors as an assistant, not as a member of staff. Businesses may give it a name and a personality, but must not make visitors believe they are talking to a person.

5. Other people's voices

The Service does not clone voices. If you use the open model weights to make a voice from someone's recording, get that person's clear permission first, and never use a voice to deceive.

6. Other people's recordings

Before you transcribe a recording of other people, make sure it was made and shared lawfully and, where the law requires it, with their consent.

7. Say when it is synthetic

When listeners could take the audio for a human recording, for example on the radio, in a message or in a video, tell them it was made with a synthetic voice.

8. Reporting misuse

If you find audio made with these voices, or an assistant, being used to deceive or harm, write to [contact email] with a link or a copy and where you found it. We may suspend the account or block access to the Service, and we cooperate with the authorities when required.

9. Speakers' wishes

If a speaker asks us to stop offering their voice, we will remove it from the Service. Assistants using that voice will be moved to another voice, and their owners informed.