OutaStory can narrate your stories in your own voice. To do that it
needs a recording of you, and from that recording it builds a voice
model — a clone.
This article is the part people most want to be told plainly: what you
are agreeing to, where the recording goes, and how to get rid of it.
The step-by-step recording guide is separate:
Record your narration
voice.
Who can make one
You have to be an author — publish a story first. If you open
/profile/narration-voice before that, it says so.
The consent screen comes first
Nothing is recorded and nothing is sent anywhere until you have read
the consent notice and pressed I Agree — Record My Voice.
In plain words, you are confirming three things:
- It is your voice, or you hold the rights to use it.
- We may build a voice model from it, to narrate stories you
publish. - We may clean the recording up first — reducing background hiss
and room noise. The cleaned-up copy is kept next to the original,
under exactly the same terms, and is deleted with it.
You can say no and simply not have the feature. Nothing else about your
account changes.
What the recording is
A short passage read aloud — about one to two minutes is right. The
screen gives you the text, so you do not have to think of anything.
A recording has to be at least 45 seconds and no longer than
five minutes. Outside that, it is rejected with a note asking you to
record again — that is a quality check, not a judgement of your voice.
You can also upload a recording you already have instead of
recording in the app.
Where your recording goes
Two places, and we would rather you knew both.
- OutaStory's own private storage. The original recording, and the
cleaned-up copy if one was made, are kept there so the clone can be
rebuilt when it is needed. - ElevenLabs, our speech provider, in the United States. They
are the ones who actually build and run the voice model. They act on
our instructions as a processor, under a contract that covers exactly
this and nothing else.
Your recording is not used to train anyone's general-purpose voice
models, and it is not offered to other people unless you separately and
explicitly ask for that — see the last section.
Why this is treated as sensitive data
A recording of your voice, and the voiceprint built from it, identify
you the way a fingerprint does. Under European data-protection law that
makes it special-category (biometric) data, and the law allows it
only on your explicit consent.
That is why there is a consent screen at all, why it is worded the way
it is, and why withdrawing consent deletes the data rather than just
switching a feature off.
It also means we do not use your recording to check that you are
who you say you are. The basic clone rests on your confirmation that the
voice is yours.
Deleting it
Open /profile/narration-voice and press Delete. You will be asked
to confirm, because it cannot be undone.
Deleting removes the recording, the cleaned-up copy and the voice model.
Narrations that were already produced with it are audio files that
already exist and are not affected — if you want one of those taken down
too, ask us at /help/contact.
Your voice data is also included in the copy of your data you can
download at any time — see
Downloading a copy of your data.
Letting other authors use your voice
That is a separate decision and a separate consent. Making a clone
for yourself never puts it in front of anyone else. If you later choose
to offer it, you are asked again, in its own screen, and you can
withdraw that permission without deleting the voice itself.
If it goes wrong
- "Recording not accepted." It was too short, too long, or too
noisy. Find a quieter room, keep a steady distance from the
microphone, and read for a full minute. - The clone never becomes ready. Give it a few minutes, then
refresh. If it is still not ready after an hour, write to us. - A narration does not sound like you. That is usually the sample,
not the model — a flat or muffled reading makes a flat, muffled clone.
Delete and record again; there is no limit on trying.