AI glossary · Voice
Speech to text
In one sentence
Speech to text (STT) is the technology that turns spoken audio into written text. It's the software component, usually a model reached through an API, that does the listening in a voicebot, a dictation tool or a transcription service, either live or on recordings.
Also known as: STT · voice to text · speech-to-text conversion · ASR
What is speech to text?
Speech to text, shortened to STT, is the part that gives a machine ears. You feed it audio, from a microphone or a phone call, and it hands back written words, often punctuated and with a timestamp for every word.
Put it to work
A voice agent that picks up the phone
From €149/mo + VAT
It answers calls 24/7, books appointments and logs every call in the CRM.
See the voice agentThree terms get mixed up a lot:
- Speech to text: the specific component or service that converts audio into text.
- Speech recognition: the technical discipline behind it, also called ASR.
- Automatic transcription: the end use, turning recordings into readable documents.
Related to this term
UTIO Voice
- 500 minutes and a number included
- Natural conversation in 26 languages
- Introduces itself as a virtual assistant
- 14-day free trial, no card needed
From €149/mo + VAT
See the voice agentHow it works
A modern STT system works in three steps:
- Prepare the audio. It slices the sound into very short frames and turns them into a numerical representation.
- Recognise. A neural network trained on many thousands of hours of speech estimates which words match those sounds, taking context into account.
- Format. It adds punctuation, capitals and digits (“twenty-three” becomes “23”).
There are two ways to use it:
| Mode | How it works | Used for |
|---|---|---|
| Streaming (real time) | Returns partial text while someone is speaking | Voice agents, live captions |
| Batch | Processes the whole recording at once | Recordings, minutes, voice notes |
In a voicebot, streaming STT is the first link in the chain and one source of latency. It has to judge quickly when the person has finished speaking, without cutting them off. The reverse step, from text to voice, is done by text to speech.
Example
A lift maintenance company in Birmingham wants its engineers to stop filling in job sheets by hand. Now, at the end of each visit, the engineer dictates what they did into their phone. An STT service turns it into text and the system drops it into the job sheet. The engineer just checks it and signs. The key detail: the STT understands technical vocabulary if you give it a list of in-house terms, such as motor models or part names.
Speech to text at UTIO
In UTIO’s voice agent, STT runs in real time throughout every call: that’s how the agent understands everyday speech, interruptions and accents without making anyone press keys. Transcription is included in the per-minute price, along with the model, the voice and the telephony, and it runs on UTIO’s own servers. Transcripts are stored in Spain and attached to each call in the CRM.
If you’re a developer, the AI API offers transcription of calls and audio to text, with the language detected.
Frequently asked questions
What is the difference between speech to text and speech recognition?
In practice they overlap. Speech recognition is the name of the research field; speech to text is the name of the function or product that delivers the text. A provider selling “STT” is selling packaged speech recognition.
Should I use real-time or batch STT?
It depends on the job. For a conversation, such as a voice agent, you need real time: text arrives while the person is still talking. For transcribing recordings, batch processing is often more accurate because the model can see the whole audio.
How is STT quality measured?
With word error rate (WER): the share of words misrecognised, missed or added compared with a reference transcript. Test it on your own audio, because it varies a lot with noise, accent and vocabulary.
Related terms
- Speech recognition
Speech recognition is the branch of artificial intelligence that lets a machine identify the words a person says and turn them into text or commands. Its technical name is ASR, short for automatic speech recognition.
- Automatic transcription
Automatic transcription is the process of turning an audio or video recording into written text without anyone typing it, using speech recognition. It usually adds punctuation, timestamps and labels showing who is speaking at each point.
- Text to speech
Text to speech (TTS) is the technology that turns written text into spoken audio. It's the component, usually a model reached through an API, that gives a voice to a phone agent, a screen reader or any audio generated from text.
- Voicebot
A voicebot is a program that holds spoken conversations: it listens to what a person says, works out what they want and answers out loud. It can run on a phone line, a website, an app or a smart speaker, and with generative AI it handles open conversation, not just menus.
- Latency
Latency is the time between making a request and starting to receive the response. In a voice agent, it's the silence from the moment the caller stops talking to the moment they hear the agent's first word, and it decides whether the call sounds natural.
Updated on