Voice Companion for 4-Year-Olds

Challenge

Building AI voice applications for children under 5 presents unique challenges — safety constraints, limited vocabulary, attention spans under 30 seconds, and the need for age-appropriate content that's both educational and engaging. Most voice AI products are designed for adults and adapted for children. This project took a child-first approach, designing the entire experience around the cognitive and emotional needs of a 4-year-old.

Approach

I started with a Product Design Requirements (PDR) document that analysed the specific needs of 4-year-old users — their language patterns, attention spans, safety requirements, and interaction preferences. Key design decisions: - Voice-only interface (no screen required) to reduce eye strain and encourage imagination - Short interaction cycles (15-30 seconds) matching attention span research - Strict content filtering with multiple safety layers - Parental controls with activity summaries - Encouragement-focused responses that build confidence The feasibility study confirmed that modern LLMs, combined with speech-to-text and text-to-speech pipelines, can deliver a compelling experience — but careful prompt engineering is needed to maintain age-appropriate responses.

Results

• Comprehensive feasibility report covering technical, safety, and UX requirements
• Product Design Requirements document detailing all user stories and acceptance criteria
• Safety-first architecture with multi-layer content filtering
• Age-appropriate interaction patterns based on child development research

Key Learnings

• Designing for children requires fundamentally different UX patterns — you can't just 'simplify' adult UIs
• Voice latency is critical for children — responses faster than 1s maintain engagement
• Safety filtering needs to be proactive, not reactive — prevention beats moderation

Gallery