Can your stored voice recordings be used to clone your voice?

Yes, in principle. A stored recording is the same kind of material as any other clip of your voice, and a voice clone needs very little of it. What decides the risk is who can reach the stored copy and how long it is kept. A recording that only you can access is not a cloning risk in itself. A copy that staff, contractors, a breach or a compromised account can reach is.

A few seconds of your voice is enough raw material

Research has shown how little audio a clone needs. Microsoft's VALL-E model could produce personalized speech from a 3-second recording of a speaker it had never heard before ¹. That was a research system, not a consumer product, but it means a stored clip does not need to be long to be useful.

Commercial tools show the same thing. Consumer Reports tested six voice cloning products and its researchers were able to easily create a voice clone from publicly available audio in four of them ². That test used public audio, not private storage. The point is that a clone does not need your cooperation or a company's files. Stored recordings are one more place the same material can come from.

A stored copy creates risk through retention and review

A stored copy matters for three reasons: it can be kept for a long time, people may be able to listen to it, or it can be exposed by a breach.

Retention is the most measurable. Microsoft says it keeps voice clips that people contribute for sampling and listening for up to two years ³. Apple says Siri and Dictation request history, which includes transcripts, may be kept for up to two years ⁴.

Review widens who can hear a clip. Microsoft says its employees and vendors working on its behalf can review contributed voice clips to build and improve speech recognition ⁵. A human listening to your audio is a different kind of access from a server that only processes it. In 2019 Google said one of its language reviewers had leaked confidential Dutch audio data ⁶. That case was about speech data, not cloning, but it shows what can happen when people can hear stored audio.

Breaches and account takeovers are general risks of any stored data. A copy that exists can be stolen, and an account that is compromised can give access to the history inside it. No setting removes this risk completely, which is why keeping less is the simplest protection.

Apple and Microsoft make keeping your audio optional

Not every service keeps your clips. Apple says dictation sent to its servers is not stored unless you opt in to Improve Siri and Dictation ⁷. Microsoft says contributing voice clips is optional, and that you can use voice typing without contributing them ⁸.

So the answer depends on the setting. A service that keeps nothing unless you opt in leaves less stored audio to reach than one that keeps clips for up to two years. On-device recognition goes further, because the speech stays on your computer and there is no provider copy to review. Yap, which we make, is an on-device dictation app for Mac and Windows, and the guide on voice data retention compares what each provider keeps.

It depends: four situations

The last row matters most for many people. Stored copies add a route, but they do not replace the public audio that already exists.

What to do

1. Check whether your audio is stored. Open the privacy or speech settings of each voice service and look for a setting that controls saving audio or using it to improve the service. Turn it off if you do not want it kept.
2. Delete history you do not need. Remove old dictation and voice history where the service lets you. Our answers on how to delete voice recordings and whether Apple keeps dictation recordings explain the steps.
3. Do not contribute clips by default. Contributing voice clips to improve a service is optional, so leave it off unless you have a reason to contribute.
4. Read the review and retention rules. Find out whether staff or vendors can listen to your clips and how long they are kept. The guide to human review of voice recordings explains the reviewer risk.
5. Limit public audio. Remove the clearest clips of yourself that you can, keep your voicemail greeting generic, and assume that any clip you post can be copied.
6. Agree a safe phrase with the people who might call you for money, as the guide on voice cloning and deepfakes explains.

What we could and could not verify

We did not find a documented case of a voice clone made from a dictation provider's stored audio. The cloning cases we checked used public audio, phone calls or other material, and that is a finding about what we could verify. It is not proof that the route is impossible. Storage is a real and avoidable risk, so the sensible answer is to limit what is kept, who can review it and how long it stays.

For the wider picture, see whether your voice can be cloned from a short clip and how to protect yourself from voice cloning scams. The overview of voice privacy covers how voice data is handled across the main services.