We let parents clone their voice for bedtime stories. Here is the safety checklist we built first.

Own-voice narration is the feature Dreamify gets the most messages about. It is also the one we were most nervous about shipping, because a cloned parent's voice inside a children's app is a trust problem long before it is a technical one. So we wrote the rules first and built to them. Here they are.

1. One purpose, written down

The recorded voice is used to narrate stories inside Dreamify. That is the whole list. It is not used to train anything shared, it is not used in marketing, it is not available to other users, and it is not a feature of any other product. We put this in the privacy policy in plain words, and we hold ourselves to it.

2. The parent records, never the child

Voice setup lives behind the parent's area of the app and reads a passage written for an adult. We do not ask for, and do not want, a child's voice. The stories are told to children, not by them.

3. Delete means delete

There is a delete button in the app's settings. Tapping it removes the voice from our systems, and narration falls back to the built-in narrators. No email to support, no waiting period, no "are you sure" maze.

4. Switching back is always one tap away

Some nights a parent wants the calm narrator instead. Some families decide the feature is not for them after trying it. Either way, switching to a built-in voice is immediate and reversible. A safety feature that is hard to use is not a safety feature.

5. The story engine has a "never mention" list

Voice is only one part of safety at bedtime. Parents can tell Dreamify topics that must never appear in a story, and the engine treats that as a hard rule. The child who is scared of thunderstorms this month will not meet one in tonight's tale.

6. Short recording, on purpose

We ask for a short reading rather than a long session. It is enough for natural narration, it respects the parent's evening, and it means we hold less audio than we otherwise would.

7. The two-sentence test

Every feature in the voice flow has to be explainable to a parent in two sentences, including what happens to their data. If we could not write those two sentences, the feature did not ship. Several did not.

Why bother

Because the feature is worth it. A child falling asleep to a story in a parent's voice on a night that parent is away is the best thing we have built. It only works if the parent trusts it completely, and trust is built from rules like these, not from a settings screen nobody reads.