A story written by people, steered by your child.
Little Wisp has two halves. Everything a child can hear is finished, checked and recorded long before play. The only live part is the moment your child answers.
Before play
Every path is written, read and approved ahead of time
An author lays out a story as a handful of fixed beats the plot must pass through, and the questions that sit between them. A writing model drafts the connective tissue between those beats. Then the author reads every line, on every path, and approves it, or changes it.
Only then is the story recorded: each line performed in a named character's voice, with music and sound effects the author chose. The result is a finite story with a few forks, and every fork has an ending.
When your child plays, nothing is written and nothing is invented. The story cannot wander somewhere no adult has read.
Fixed beats
The moments the story must pass through. The author writes them; nothing may move or skip them.
Questions with two to four exits
Each exit is a labelled choice. The author also writes the gentle re-ask for when an answer is unclear, and the reply for when a child picks something they already tried.
Approved, then recorded
Any edit clears the approval. A story is published only when the whole manuscript has been read and approved again.
During play
The ten seconds after your child speaks
This is the only live part of Little Wisp, and it is deliberately small.
- 1
The story asks
At a question, the story stops and the microphone opens. The browser asks for microphone permission once, the first time.
- 2
Your child answers
They just talk. When they go quiet for a moment, the answer is over. If they say nothing for a few seconds, the story notices that too.
- 3
Words, then a choice
The short clip is turned into words by a speech-to-text service, and the words are matched against the two-to-four choices the story offers. That is the only thing the matcher may answer: one of those choices, or none.
- 4
Then it is gone
Only the choice taken is saved, so the story can resume tomorrow. The recording and the words are not stored anywhere.
If the answer is unclear
The story asks again, in words the author wrote for that moment. After two tries without a usable answer, the story picks a path itself and carries on. Nobody gets stuck.
If the microphone is blocked
A grown-up can type what the child said, or tap the choice. The story continues either way.
If a question comes round again
Some stories loop back. A choice already taken plays its authored “you already did that” reply, then asks again.
Plain about the technology
Where AI is used, and where it is not
Used, under a human
- Drafting story text between the author's fixed beats, at writing time
- Performing the approved lines as speech, at recording time
- Turning your child's answer into words, and matching it to an offered choice
Never
- Generating story text, voices or sounds while a child is playing
- Holding an open-ended conversation with a child
- Keeping recordings or transcripts, or building a profile of a child
The companies that provide these services are named in our privacy notice.
What you need
Not much
A browser
A laptop or tablet with a microphone. Little Wisp is in early access and runs in the browser today.
A grown-up
To pick the story and press play. Your account, your consent, your call.
A quiet-ish room
The story listens for your child's voice. It copes with a bit of background noise; a running tap is pushing it.
Recommended for children aged [INTENDED AGE RANGE, e.g. 4–8].