TTS (Text-to-Speech)
Converting text to natural-sounding audio — OpenAI TTS, ElevenLabs.
Simple explanation
TTS (text-to-speech) converts text into spoken audio. Modern TTS produces near-human voices with emotion and pace control. It's the output stage of voice assistants.
Engineering explanation
TTS models: OpenAI TTS (alloy, nova, echo voices), ElevenLabs (highly realistic, voice cloning), PlayHT. Production: stream TTS at sentence boundaries for low latency, cache common responses, maintain consistent voice identity, support barge-in (interrupt when user speaks).
Example
LLM response: 'The weather is 18°C.' → TTS → audio file/stream → user hears spoken response.
When to use it
Voice assistants, accessibility, audiobooks, IVR systems.
Common misconception
TTS is NOT just 'reading text aloud'. Modern TTS adds emotion, pacing and natural prosody. The voice identity must be consistent across sessions.