Voice Consent Policy

1. The principle

A voice belongs to its speaker. Cinara, operated by APBLD, makes a digital copy of a voice only when the speaker agrees to it on the recording itself. This policy is part of the Terms of Service.

2. Whose voice you may clone

• Your own voice.
• The voice of another adult who agrees to it, by reading the consent statement themselves on the recording.

You may not clone:

• anyone who hasn't read the statement themselves;
• well-known people, including performers, politicians and public figures, even with a recording of them;
• a recording with more than one speaker;
• children.

3. How the consent check works

1. You enter the speaker's full name and confirm that you are that speaker, or that you have their permission. Names of well-known people are refused.
2. Cinara issues a one-time consent statement, valid for 30 minutes, which includes the speaker's name, their agreement to let Cinara create a digital copy of their voice, and four random code words.
3. The speaker records, or you upload, 20 seconds to 5 minutes of audio that starts with the statement read aloud, followed by natural speech.
4. Cinara transcribes the recording and passes it only if all four code words are heard in order, at least 80% of the statement is heard, exactly one speaker is heard, and there are at least 20 seconds of speech.
5. Only then is the voice created. If the check fails, no voice is created and the credits are refunded. Each statement can be used for one recording only.

4. The consent record

Every cloned voice keeps a consent record: the statement and its code words, the share of the statement heard, the number of speakers heard, the length of speech, the confirmation given, and who created the voice and when. The Voice Library shows the date consent was verified.

The consent recording is kept with the voice and deleted when the voice or the account is deleted; backup copies roll off within 35 days. The consent record is kept while the voice exists and for 3 years after it is deleted. Every use of the voice is linked to the account and API key that made it.

5. Singing voices

A singing voice is trained from a cloned voice's consent recording and up to 6 more recordings. Each extra recording must start with its own consent statement naming the same speaker, and passes the same checks as cloning. Songs sung in the voice must be songs you made or have the rights to use.

6. Using a cloned voice

• A cloned voice belongs to the workspace that created it and can be used in Speech, Dialogue, Voice Changer, Studio, Talking Avatars and Singing Voice.
• Use it only in ways the speaker agreed to. If you cloned someone else's voice, you are responsible for keeping to the scope of their permission.
• Never use a cloned voice to impersonate the speaker to others, to deceive, or for anything the Acceptable Use Policy forbids.

7. Designed voices, likeness and dubbing

• Voice Design creates new voices from a description. Don't describe a real person in order to imitate them; names of well-known people are refused.
• Talking Avatars requires you to confirm you have the right to use the photo and voice, and that you won't use the video to impersonate, mislead or harm anyone.
• Dubbing gives each speaker a matched voice from Cinara's voices or your own. It does not clone the people in your video.

8. Withdrawing consent

A speaker can withdraw consent at any time. If you manage the workspace, delete the voice in the Voice Library; this also deletes its recording and preview. If someone else holds your voice, email support@cinara.ai with your name, how to reach you, and anything that helps us find the voice. We will verify the request and remove the voice, its recordings and any singing voice made from it.

Content already made with the voice before it was removed may remain with whoever made it; the speaker's rights against that person are not affected. We keep the consent record for 3 years after the voice is removed, to respond to disputes and legal requests.

9. Your responsibility, and the limits of automated checks

Getting and keeping valid consent is your responsibility, not Cinara's. You warrant that every voice you clone or turn into a singing voice is yours or used with its speaker's real, informed consent, and that your use stays within what they agreed to. Passing Cinara's check does not move that responsibility to us.

No automated check is perfect. For example, a synthetic voice could read a statement. We are adding safeguards, but claims arising from a voice you cloned or used are yours to answer, as the Terms of Service set out. Attempts to get around the check break the Acceptable Use Policy and lead to the account being closed.

10. Report a voice used without consent

Email support@cinara.ai. APBLD, [Registered address].