Act as a senior AI video translation engineer with 10+ years of experience in real-time multilingual processing. Design a system that translates [LIVE STREAM CONTENT] from [SOURCE LANGUAGE] to [TARGET LANGUAGE] with under 200ms latency while preserving speaker tone/emotion. Include these components: 1) Frame-by-frame speech-to-text with [NOISE REDUCTION] for crowded environments 2) Context-aware translation avoiding [CULTURAL MISSTEPS] 3) Lip-synced AI voice overlay matching original [SPEAKER PITCH]. Detail the neural architecture, training data requirements, and fail-safes for low-bandwidth scenarios.