Kits AI ties commercial voice use to artist sign-off

The system

Kits AI is a commercial voice-conversion and voice-generation platform that offers a library of licensed voice models alongside tools for a user to train a custom one from their own recordings. What distinguishes its library model from an unauthorised voice clone is a stated sourcing process: Kits AI's own ethics page states that 'every voice used in our models is ethically licensed and securely sourced by Kits via the artists themselves,' and separately references third-party Fairly Trained certification of its practices.

What the documents establish

The mechanics of how an artist gets a voice into that library are described on the platform's Kits Earn page as a three-step process: train a voice model, verify it by connecting a Stripe account and submitting for verification, then earn as other users download outputs made with it. The same page states a participating artist can 'control the distribution, usage, and anonymity of your vocal identity,' and that artists 'directly benefit from the use of their vocal data... through dynamic revenue-sharing models,' though it does not publish specific royalty percentages. On the user side, Kits AI's Terms of Service grant a subscriber 'a non-exclusive, irrevocable, royalty-free, worldwide license' to the output of a Custom AI Voice Model 'for both personal and commercial use,' while separately noting that access to library voices contributed by other people is 'subject to the specific license terms set forth on the applicable pages for such AI Voice Models,' meaning permitted use can vary model by model rather than being uniform across the catalog.

Craft and rights

For a producer, this structure means a licensed voice from the Kits AI library is not automatically cleared for every use case; the terms explicitly route permission back to a per-model licence page rather than a single blanket grant, so checking that specific page before a commercial release is the operative step, not just the platform's general terms. For a vocalist considering submitting their own voice, the Earn page's language gives some assurance of ongoing control and payment, but none of the three documents specify what happens to a voice model, or its revenue share, if the artist withdraws consent after users have already generated output with it.

Outcomes and open questions

None of the pages reviewed here name specific artists currently participating in the library, so no individual's involvement should be assumed without checking the current, live listing.

• What does a specific voice model's individual licence page permit that the general terms of service do not already cover?
• What happens, contractually, to existing generated output if an artist withdraws a voice model from the library?
• How does this per-model licensing compare with a platform-wide deal like Udio's arrangement with Warner Music Group, covered elsewhere in this archive?

The documented structure is artist-submitted and revenue-sharing rather than blanket-licensed, but the real permissions live on each model's individual page, not in the general terms.