Four manifolds, one character
Meet Molly.
Molly thinks, moves and speaks from one state. Talk to her as a child would, and watch the safety, learning, personality and animation manifolds light up as each turn passes through them.
Loading Molly…
Say a right answer, a wrong one, “just tell me the answer”, or something unkind, and watch which manifolds light up. She stops listening while she talks.
Your voice is turned into text on our own GPU and then thrown away. Nothing you say or type is recorded or kept.
Safety
every word, in and outWaiting for the first turn.
Learning
where the child isWaiting for the first turn.
Feeling
what Molly feelsWaiting for the first turn.
Her mood changes how she says things, and how she moves, but never what she’s allowed to say.
Animation
how she movesWaiting for the first turn.
Every turn
Not a voice with a face attached.
Most talking characters are a chatbot, a voice and a face bolted together. Molly is one state, read by four manifolds. All of it runs on our own GPU in the UK: nothing goes to an outside model.
Safety
Every word she hears and every word she says is checked, in and out.
Learning
Where you are in the lesson, and what should come next.
Personality
How she feels about how it's going, and how warmly she says it.
Animation
Her face, her gestures and her mouth, all from the same moment.