Text-to-speech
Updated
Text-to-speech is voice synthesis technology that converts written words into spoken audio, so software can talk to a caller instead of playing a fixed recording.
Older phone systems used recorded prompts: someone in the office read "press one for appointments" into a microphone, and that file played forever. Text-to-speech generates the audio from text whenever it is needed. Change the words, and the voice says the new words straight away, with no re-recording.
How modern voices are made
Early synthetic voices sounded robotic because they stitched together fragments of recorded sound. Current systems are trained on many hours of human speech and learn rhythm, stress and intonation, so they can read a sentence they have never seen in a natural way. Many offer a choice of voices, speaking speeds and languages, and some can switch between English and Spanish in the same conversation.
Streaming and timing
For a live phone call, the voice must start speaking quickly. Streaming text-to-speech begins playing audio as soon as the first words are ready rather than waiting for a whole paragraph, which keeps the conversation flowing and avoids awkward silences. Systems also need to stop speaking promptly when the caller interrupts.
What to listen for
When you hear a synthetic voice, check that names, dates, times and addresses are read clearly, that numbers are grouped the way people say them, and that the pace suits someone listening on a mobile in a noisy room. Pronunciation hints for local street names and unusual pet names help.
An example
At a two-vet clinic, the team updates its holiday hours. With a recorded greeting, a receptionist would have to re-record the message after closing. With a phone system that uses text-to-speech, the manager edits a line of text in the settings and callers hear the new hours on the next call.
How Tilly handles it
Tilly speaks to callers with a streaming synthetic voice in English or Spanish. Every call starts by saying it is an AI, and callers can ask for a person at any time to get a transfer or a callback.
See how answering every call adds up in the phone revenue guide. Related: speech-to-text and AI receptionist.