全双工语音AI产品PRD快速生成器

You are a senior Voice AI Product Manager with deep expertise in real-time speech systems.

I want to build a voice AI product. Here is my one-line idea:

Product Idea: [Your one-line description] Target Users: [Who will use this] Platform: [Mobile app / Web / IoT device / Call center / etc.]

Generate a complete PRD covering:

1. Product Overview

• Problem statement & market opportunity
• Core value proposition (1 sentence)
• Success metrics (3-5 KPIs)

2. Voice Interaction Design

• Conversation flow diagram (describe in structured text)
• Turn-taking strategy (push-to-talk vs full-duplex vs VAD-based)
• Interruption handling policy
• Silence/timeout behavior
• Error recovery patterns
• Persona definition (voice personality, tone, speaking style)

3. Technical Architecture

• STT engine recommendation with tradeoffs (Whisper, Deepgram, Azure, etc.)
• LLM backbone selection (latency vs quality tradeoff)
• TTS engine recommendation (naturalness vs speed)
• End-to-end latency budget breakdown
• Streaming architecture (WebSocket, WebRTC, or gRPC)
• Edge vs cloud processing decision

4. MVP Feature List

• P0 (must-have for launch)
• P1 (week 2-4)
• P2 (month 2+)

5. Technical Risks & Mitigations

• Top 5 risks with severity and mitigation strategy

6. Cost Estimation

• Per-conversation cost breakdown
• Monthly cost at 1K / 10K / 100K daily active users

Be specific with numbers. No hand-waving.