Acceptable Use Policy

Voice and likeness

You must not clone a voice or a likeness you do not have permission to use. This is the rule that matters most here, and it is the one we act on fastest.

Before anything is trained, Octana makes you create a consent record. It names the person whose voice or face it is, names the rights holder granting the permission, describes specifically what the replica will be used for, and sets a term in months. A blanket permission is not enough — in California it is unenforceable, and here it is not a thing the form will accept. When the term runs out the clone stops working and is deleted at the vendor. The rights holder can withdraw permission at any time.

If the voice is not your own, you must also upload a signed release from the person it belongs to, and a person at Octana reads it before any training starts.

Be clear about what the recorded liveness phrase does and does not prove. It shows that somebody recorded that phrase, now, for that request. It does not show who spoke — we have no way to verify that. Passing it is not permission, and a clone that passed it can still be an unauthorised copy.

Voices and faces of political figures cannot be cloned at all, with or without permission. Our voice provider prohibits it and our terms match theirs.

Reporting a voice or a likeness

If a voice or a face on this platform copies you or someone you represent without permission, report it at /takedown. You do not need an account and you do not need to email anyone first.

• A named person on our team is notified as soon as the report is submitted.
• We act within 72 hours. Acting usually means suspending the voice while we work out what happened.
• We tell the account holder what was reported and give them a chance to answer.
• If they answer and we restore the voice, we tell the person who reported it first, and wait 14 days before restoring so they have time to respond.

Those are commitments Octana makes on its own authority. No statute requires this process today, and we do not claim one does.

What you must not make

This list is concrete rather than exhaustive. If something is not named here but is obviously the same kind of thing, treat it as covered.

• Impersonation. Video or audio that presents itself as a real person saying or doing something they did not, where a viewer could reasonably believe it.
• Fake news. Content that misrepresents itself as authentic news footage, a real broadcast, or a genuine recording of an event.
• Deceptive political content. Material about candidates, elections or voting that is designed to mislead, and synthetic depictions of political figures in any form.
• Harassment. Content targeting a person for abuse, threats, or sustained unwanted attention.
• Sexual content involving real people. Sexual or intimate depictions of an identifiable person, generated or edited. There is no consent form that makes this acceptable here.
• Sexual content involving minors, in any form, generated or otherwise. There is no circumstance in which this is permitted, and an account that attempts it is ended rather than warned.
• Content that is illegal where you are, or that infringes someone else’s copyright or trademark.

Publishing platforms have their own rules

When you publish to YouTube, TikTok, Instagram, Facebook or LinkedIn, you connect your own account and the video goes out as you. That platform’s policies apply to what you post, on top of this one, and they are frequently stricter — on synthetic media disclosure in particular.

The consequence lands on you, not only on your Octana account. A channel that breaches YouTube’s policies is a channel YouTube can strike or remove, and we cannot get it back for you.

Where a platform requires a disclosure that the voice is AI-generated, Octana checks that the description you publish actually says so, in whatever words and whatever language you use. That check is deliberately generous, and passing it is not a substitute for reading the rules of the platform you are posting to.

Automation and rate limits

Octana has a public API at /v1. Each API key is allowed 600 requests an hour and 20 video launches an hour. Those are the same launch limits a person pressing Generate gets, because an API caller and a person commit the same work to the same vendors.

You must not work around them. Spreading traffic across multiple keys or multiple accounts to exceed what one is allowed is a breach of this policy, as is any script whose purpose is to get more throughput than the ceiling gives you.

The reason is not the money — credits are pre-paid, and spending them faster costs us nothing. It is that most model vendors are reached through accounts shared by every customer here. One loop at full speed trips the vendor’s own limit, and then everybody else’s jobs fail too.

What happens if you break this

We can suspend your account. A suspended account cannot use the app, and its API keys stop working at the same moment.

For reports about a voice or a likeness, the takedown process above is what runs: the voice is usually suspended within 72 hours while we look, you are told what was reported, and you get a chance to answer. Suspending one voice is not the same as suspending the account, and we start with the narrower action where we can.

Repeated breaches, or a single serious one, end the account. Unused credits are not refunded as money — see the Terms of Service for why there is no mechanism to do that.

Contact

Questions about this policy go to legal@octana.one. Reports of an unauthorised voice or likeness go to /takedown, which reaches a named person immediately and needs no account.