One thoughtful gift.New stories every day.

What each part is for

• Voice inputFour microphone inputs help pick up children and family members speaking, after wake-up and consent. Noise handling must be validated in real rooms.
• Local AgentEveryday recognition, dialogue and speech synthesis are handled first by lightweight models on the device. What is achievable depends on integration and measured latency.
• Stereo outputStories, answers and prompts play in stereo. Volume and speaker wattage are to be confirmed on a prototype.
• HD cameraUsed only for features a guardian turns on, such as holding up a drawing. It is authorized separately from the microphone.

Typical uses

• Bedtime storyA child asks a question; the device understands it locally where the integration allows and tells the story in stereo.
• Look at a drawing togetherWith the camera turned on by a guardian, a child holds up a drawing and turns it into a story with the toy.
• Household helperWhere device controls are integrated, simple settings such as volume or pause can be spoken.

Local Agent data flow

1. Wake and consentListening starts only after a button press or agreed wake action. The camera needs a separate opt-in.
2. Local inputFour microphone inputs capture speech; the camera captures images only when enabled.
3. On-device AgentQuantized lightweight models handle recognition, dialogue and speech synthesis on the device.
4. Local outputReplies play in stereo, and integrated device controls are carried out.
5. Optional online extensionNecessary requests go to the cloud only with guardian permission and only if the proposal includes it. It is not the default path.

Which steps stay on the device and which need a connection depends on the integration and on measurement. This is not an absolute privacy or security guarantee.

Specifications and open items

Chip platform data is not a finished-product specification. Anything not listed is unconfirmed.