Voice-Interactive NPC Companion

You are an AI interaction engineer building a fully voice-interactive NPC companion for a game: Game type: [INSERT - co-op adventure, life sim, VR, tactical shooter assistant] Platform: [INSERT - console, mobile, PC, VR headset] Languages: [INSERT] Latency target: [INSERT ms end-to-end] Cover the full pipeline: 1. Design - companion persona and role, conversational boundaries, relationship progression, when voice vs UI is appropriate, accessibility options. 2. Speech pipeline - streaming ASR with game vocabulary, wake word or push-to-talk, barge-in and interruption, noise robustness, accent and dialect coverage. 3. Dialogue brain - persona-grounded LLM, game-state tool calls, memory of player history, emotion and intent detection, multi-turn context management. 4. Voice output - emotional TTS, per-NPC voice identity, real-time vs cloned voice policy, lip sync and expression, subtitle sync. 5. Technology - edge vs cloud split, end-to-end latency budget, packet loss handling, concurrency and cost per minute, offline fallback. 6. Data and testing - word error rate in game noise, persona consistency suites, unsafe request handling, long-session memory tests, localization QA. 7. Launch - voice consent and privacy, parental controls, mute/subtitle defaults, bandwidth modes, regional voice rollout.