What you are actually selling
You are not selling your voice. You are selling a licence: permission for a named company to use specific recordings, for a specific purpose, for a specific time. That distinction is the whole game. Keep it, and your voice remains an asset you can licence again and again. Lose it - through a perpetual, irrevocable, sublicensable grant - and you have handed over your voice identity for a one-off fee.
Fair terms look like this
- A named client and purpose. "Training the TTS voice of product X" - not "any current or future technology".
- A defined term. Licences should end. Renewal is a new conversation and a new fee.
- No resale. Your recordings train the named project and are not repackaged into datasets you never agreed to.
- Payment before scope. The fee reflects how widely the voice will be used - and you see both before recording.
- The right to be forgotten. Under UK and EU data protection law, your voice recordings are personal data. Deletion routes should exist and be written down.
Who is in demand
Everyone assumes this market only wants polished announcer voices. The opposite is true: models need diversity. Regional accents, second-language English, older voices, younger adult voices, natural conversational delivery - all of it has value. If you speak a language or carry an accent that is under-represented in speech data, you are precisely who AI teams struggle to find.
The route in is our two-minute application on the AI voice jobs page. If you want the deeper picture of how sessions and licensing work, read voice acting for AI training, and see the standards we hold every client to on our ethics and consent page.
