Text-to-speech (TTS) turns text into synthesized speech for applications such as accessibility, voice assistants, narration and customer support. Systems differ in naturalness, pronunciation control, latency, supported languages and whether they can stream audio as it is generated.
Typical uses include screen readers, audio versions of articles, in-app guidance and voice agents that talk to callers. For an interactive AI agent, delay matters: audio that starts streaming sooner feels more natural than waiting for a whole paragraph, so watch inference latency. Spell out acronyms, numbers and dates if the voice mispronounces them.
Check pronunciation of names and technical terms, and obtain the required consent and rights before cloning or imitating a real person’s voice. Related: Speech Recognition, Speech-to-Text.