Glossary
Speech-to-text
In short
Speech-to-text, also called speech recognition or transcription, is technology that turns spoken words into written text, either live or from a recording.
In a phone agent, speech-to-text is the ears. It turns what the caller says into text the AI can understand, fast enough to keep the conversation natural, and the same technology produces the written transcript of every call.
Accuracy depends on audio quality, accents, background noise and specialist words such as product names, street names or medical terms. Good systems can be given the words your business uses most, so they are recognised correctly.
Its counterpart is text-to-speech, which gives the agent its voice.
A worked example
A caller says "I'm on Kirkcaldy Road, the house with the blue door." A general speech-to-text system may mishear the street name; one given the business's service-area street names gets it right, and the booking lands at the right address.
When it matters, and when it doesn't
Accuracy matters most for names, numbers and addresses — the details that end up in a booking. For general conversation, today's systems are reliable enough that most callers never notice them.
Common questions
How accurate is speech-to-text?
For clear audio in common languages, very accurate; noisy lines, strong accents and unusual words still cause errors, which is why good agents repeat back key details.
Does it work with accents?
Modern systems handle a wide range of accents, and accuracy improves when the system is told the words your callers commonly use.
Is the transcript stored?
That depends on the provider and your settings; good systems let you choose how long transcripts are kept.