Before cloning a voice: What should a broadcaster agree with the speaker?
A presenter agrees to record an introduction for a campus radio show. A few weeks later, the team wants to generate new announcements using an AI version of the same voice. The original recording agreement does not necessarily settle the details of that new use.
That distinction sits at the heart of voice-cloning consent. Does the speaker understand where their voice will be used, what it may be made to say and who will have access to the sample? A useful agreement begins with those specific questions, rather than assuming that a single recording authorises every future project.
One programme is not every programme
Imagine the presenter agrees to synthetic introductions for a particular show. Advertising, another programme or a new language may raise a different question about scope. The team and speaker can discuss who approves scripts, where the audio will be distributed and how long production is intended to continue.
If the sample is uploaded to a third-party service, the speaker should understand that provider's role. The team should examine the service's handling and usage terms as well. Permission to use a voice does not automatically clear rights in music or scripts used alongside it.
What happens if permission is withdrawn?
Stopping future generation is one task. Handling released episodes, backups and files held by an external provider can be another. A station may be able to delete a local sample without having the same control over every copy held elsewhere.
Discuss retention, access and the process for a withdrawal request before the project begins. The agreement should reflect what the team can actually do under its technology and contracts, rather than promise a deletion outcome it cannot verify.
Listen while exploring this story
Tune into campus radio with non-stop independent music and culture.
Privacy information and audience disclosure serve different purposes
Where personal data is processed, informing the person and determining the legal basis for processing are separate questions. Türkiye's Personal Data Protection Authority guidance describes information such as purposes, transfers and individual rights. The appropriate legal basis depends on the actual processing and should be assessed by the responsible team; a single checkbox does not settle every obligation.
Telling the audience about synthetic audio is a separate publication decision. Apple Podcasts' content guidelines require prominent disclosure of AI-generated audio or video in the content and metadata of each relevant episode and show. Other platforms may have different terms.
For example, if the fictional presenter's cloned voice opens a podcast episode, the team could include a clear spoken explanation in the episode and describe the use in the episode and show metadata. This illustrates Apple's requirement to disclose in both content and metadata. Agreement with the speaker and disclosure to the audience answer different questions; one does not replace the other.
Make the intended use understandable
A clear disclosure can help listeners distinguish generated speech from a recording made by the speaker. The speaker, in turn, needs a clear picture of the project before agreeing to take part.
For the campus presenter, a useful opening conversation names the programme, the kind of scripts, the distribution channels and the proposed duration. From there, the organisation's responsible team can develop an agreement suited to the actual circumstances.
