For people who have lost their voice, emotional voice systems must infer emotional intent from residual muscle signals, facial movements, or text input in the absence of real-time acoustic input, placing far higher demands on latency and accuracy than typical TTS scenarios.