What is Text-to-Speech (TTS)?
Text-to-Speech: Technology that turns written text into spoken audio, now natural enough for live phone conversations when streamed with low latency.
Text-to-speech converts text into audio. Modern neural TTS from providers like ElevenLabs, Cartesia, OpenAI, Google, and others produces voices with natural rhythm and emotion, and many support voice cloning and fine control over pronunciation. For live calls, TTS streams audio: it starts speaking the first words while the rest of the sentence is still being generated.
The key numbers are time to first byte (how quickly audio starts after text arrives), naturalness, and pronunciation accuracy on names, numbers, and brand terms. Most systems support pronunciation dictionaries or markup to fix words they get wrong. Cost is usually per character or per minute of audio and is a meaningful share of a voice agent's running cost.
Why it matters when you build or buy AI
Callers judge a voice agent within seconds, and the voice is a big part of that judgment. Pick a voice that fits your brand, test it on your real phrases (addresses, prices, product names), and keep responses short. Long, written-style answers sound unnatural when spoken, so prompts for voice agents should ask for brief, spoken-style replies.
Example
A restaurant's voice agent was saying "two to four PM" as "two hyphen four P M." Adding a pronunciation rule and asking for spoken-style output fixed it.
Related terms
More in Voice AI
Book 20 minutes, or send a brief
A former CTO replies. You leave with a written plan and a fixed estimate.
Prefer LinkedIn? Connect with Harish.