Talent — Privacy Policy
Talent is a native Mac voice training and voice library app, part of the PolySpeak voice-production suite. Because Talent trains AI voice models from recordings of real voices, this policy is deliberately direct about what the app does, what stays on your device, and where your responsibility as a user begins.
Summary
Talent makes no network requests of any kind. Voice training and voice library management run entirely on your Mac using a local voice pipeline. Talent does use your microphone to record training samples, but only while you are actively recording, and only with your explicit permission. The app has no server, no account, and no analytics. Talent also asks you to attest to consent before training a voice — see "Consent Is Your Responsibility" below, which we consider the most important section of this policy.
What Talent Accesses
- Microphone — used for recording training samples. This requires the standard macOS microphone permission prompt, and the microphone is only active while you are actively recording
- Audio files you explicitly import as training source material
- The local voice-library folder, where trained voice bundles are stored — model files, metadata, consent records, and resume/credit history
Talent does not access your camera, contacts, location, photos, or calendar.
What Leaves Your Device
Nothing
Voice training runs a local Python/RVC pipeline via subprocess entirely on your Mac. The recording used to train a voice, and the trained model itself, are never uploaded anywhere — never sent to DataIsland, and never sent to any third party. Everything Talent produces stays in your local voice library.
Consent Is Your Responsibility
Talent lets you train an AI voice model from a recording of someone's voice. This is powerful, and it can be misused. You must have the consent of the person whose voice you are recording and training a model from — or use only your own voice.
Talent includes a consent-attestation step at training time, and the attestation is stored alongside the resulting voice bundle in your local voice library. But we want to be direct rather than defensive about the limits of that feature: DataIsland does not and cannot verify real-world consent. The attestation records that you affirmed you had consent at the time of training — it is not proof, and Talent has no way to confirm who actually spoke in a recording or whether that person actually agreed to it. That responsibility is yours, and it does not transfer to DataIsland by virtue of the attestation existing.
If you are training a voice model from someone else's recording — a collaborator, a voice actor, an interview subject, anyone — get their clear consent first, and keep your own record of it. See also the Terms of Use, which addresses the consequences of misusing voice-cloning technology.
What Is Stored Locally
- The entire voice library — trained voice models, metadata, consent attestations, and resume/credit history, all in a local folder on your Mac
None of this data is synced to iCloud or any server — it stays in your local voice library.
What Talent Does Not Do
- Does not transmit any voice data or recordings anywhere
- Does not require an account or login
- Does not collect analytics, usage statistics, or crash reports
- Does not access your microphone except while you are actively recording
- Does not verify real-world consent — see above
- Does not share data with third parties
Children's Privacy
Talent is not directed at children under 13. We do not knowingly collect information from children.
Changes to This Policy
If we make material changes to this policy, we will update the "Last updated" date above. Continued use of Talent after changes constitutes acceptance of the updated policy.
Contact
Questions about this policy:
support@dataisland.com
DataIsland Software LLC