How StoryTime Together creates & reviews stories

Every story your child reads on StoryTime Together starts as AI-generated text, personalised to your child's name, age, interests, and reading level. We're direct about that — surveys of parents (including a 2025 study from NC State) show that families accept AI in children's stories when the disclosure is honest and the safety story is clear. So here's ours.

What the badges mean

• AI-generated — the story was written by an AI model from your child's profile and the theme you chose. Filters strip out scary content, age-inappropriate vocabulary, and anything that doesn't fit a calm bedtime tone. This is the default badge on every freshly-generated story.
• Reviewed by our editorial team — a human editor with a children's-content background has read the story end-to-end and confirmed it meets our editorial guidelines. We're currently building this review programme; the badge will start appearing once an editor is in place. We'll never apply it retroactively to stories that haven't actually been reviewed.
• Flagged for review — the story tripped one of our automated guardrails or was flagged by a reader. It's pulled from your child's library until an editor checks it. You can flag any story yourself from the reader.

What we won't do

• Pretend a story is human-written when it isn't. The AI badge is on every story until a human has actually reviewed it.
• Train models on your child's name, photos, or your reading patterns. Your family's data isn't our training data.
• Show stories that haven't passed our automated content filters — frightening imagery, sad endings, scary cliffhangers, or anything off-register for bedtime.

AI narration + voice cloning (Premium+)

Stock narration uses OpenAI's TTS-1-HD (Premium) and ElevenLabs Multilingual v2 (Premium+). On Premium+, you can also clone your own voice — about 90 seconds of recording — so your child hears stories in your voice when you can't be there to read.

• Consent attestation — before recording, you confirm in writing that the voice is your own (or belongs to someone in your care who has given explicit consent), and that you consent to the AI model being created for personal narration only. The attestation timestamp + IP are recorded in our audit log.
• Speaker verification — after the training passages, we ask you to read a randomly-generated phrase. We use ElevenLabs Speaker Identification to confirm the captcha speaker matches the training speaker. Below the threshold, the clone is rejected. This stops anyone cloning a voice that isn't theirs.
• AI-provenance disclosure on every MP3 — every chapter narrated with a cloned voice has an AI-disclosure ID3 tag embedded in the audio file (provider, timestamp, voice label, friendly disclosure sentence). Any standard MP3 reader (iTunes, VLC, ffprobe) surfaces it. This is our v1 of the C2PA provenance standard — a fully-signed manifest is on the roadmap.
• One-tap deletion — delete a voice in Profile → Your voices and we soft-delete locally plus hard-delete the model on ElevenLabs. Stories already narrated stay safe in your library; they were rendered to MP3 and don't need the live voice model anymore.

Premium+ families get 3 voice slots. The training audio is used only to build the voice model — never resold, never used to train any public model, never shared with any third party other than ElevenLabs (who provide the model service under our agreement).

Got a story that didn't feel right?

Tap the flag icon in the reader and we'll pull it from your library until an editor checks it. We read every flag — the bar for "is this OK for a seven-year-old at bedtime" is set by parents, not by us.