AI for Kids That Responds
Faster and Reads the Mood
Kidodo is edge0’s in-house AI hardware brand for early learning at home, maker of the AI companion robot and the Kouyubao spoken-English translator. Both products run on one hybrid edge-cloud voice stack with an on-device multimodal model: read-along and high-frequency commands complete on-device, open conversation goes to the cloud on demand, and emotion tags plus device commands are returned in the same pass.
Early Learning
Voice conversation in a product for kids has to be fast and stable, and it has to hear how the child feels in the moment. edge0’s hybrid edge-cloud pipeline lets the companion robot and Kouyubao answer instantly in high-frequency interaction, keep full conversations on open topics, and put emotion signals straight into how we encourage and what the device does next.
Companionship Needs Instant Replies,
Learning Needs Full Intelligence
The on-device side guarantees speed and stability for high-frequency interaction; the cloud carries open QA, long conversations and knowledge content. One voice pipeline schedules both paths.
Kids Don’t Wait for the System to Think
Read-along, question-and-answer and device control happen inside continuous conversation. A pause caused by network jitter breaks a child’s attention and willingness to speak.
Responses Need Emotional Understanding
The same sentence can carry joy, hesitation or frustration. Recognizing words alone makes companionship and learning feedback feel flat.
Two Task Types Need Two Paths
Device control wants instant stability; open dialogue and knowledge QA need cloud capability. A single all-cloud pipeline cannot serve both experience and cost.
One Voice Foundation for
Companionship, Learning and Translation
The AI companion robot and Kouyubao reuse the same recognition, understanding, emotion judgement and routing capabilities, then connect to their own product features through a standardized command interface.
One Entry Point, Routed to
the Right Capability
On-device handles high-frequency commands, core learning flows and emotional cues; the cloud adds knowledge, translation and open dialogue. Everything returns as one voice reply plus actions.
Tuned for Children’s Speech
Recognition is adapted to children’s pitch, speech rate, incomplete articulation and household noise, keeping real parent-child scenes stable.
Unified Edge-Cloud Routing
Wake-word, device control and fixed learning flows stay on-device; open QA and long conversations route to the cloud by intent.
Emotion Signals Enter the Dialogue
Emotion tags are emitted alongside content understanding, so the companion robot can adjust tone, encouragement and pacing.
Standardized Command Dispatch
Natural expressions become standard commands for playback, read-along, translation, volume and content switching — reused by both products.