Speech-to-text
Speech-to-text, also called automatic speech recognition (ASR), is software that converts spoken audio into written text. It powers transcription, captions, voice commands and the listening half of voice agents.
Also called: ASR, automatic speech recognition, transcription, STT
By Divya Sharma · Checked Nov 15, 2026 · How we review
A speech-to-text model takes audio, usually split into short frames, and outputs the words it hears, often with timestamps. It is the first step in meeting notes, podcast transcripts, subtitles, call analytics and any voice agent.
How it works
Modern ASR models are neural networks trained on large amounts of paired audio and text. OpenAI's Whisper is a well-known example: it was trained on 680,000 hours of audio and transcripts from the web, and it handles many languages without task-specific fine-tuning. You can run it yourself with openai-whisper or smaller variants such as Whisper large v3 turbo.
What to compare
Word error rate on audio like yours, speed, streaming support for live use, language coverage, punctuation and price per minute. Benchmark scores on clean audio can look much better than results on noisy phone calls.
Related terms
ASR tells you what was said. Speaker diarization tells you who said it, and many transcription tools combine the two. The reverse direction is text-to-speech.
Limits
Accents, crosstalk, background noise and domain jargon raise error rates. Some models also invent words during silence or music, a form of hallucination.
Related terms
- Text-to-speech
- Text-to-speech (TTS) is software that turns written text into spoken audio. Modern TTS uses neural networks that predict the sound of speech from the text, then render it as a waveform.
- Speaker diarization
- Speaker diarization is the task of splitting an audio recording by speaker, labelling each stretch of speech with who said it. It answers the question "who spoke when".
- Voice agent
- A voice agent is an AI agent that holds a spoken conversation in real time, listening to a person, deciding what to say or do, and answering in a synthetic voice, often over a phone line or in an app.
- Speech-to-speech model
- A speech-to-speech model takes spoken audio in and produces spoken audio out directly, without first converting the speech to text and back. It is used for live voice conversation and translation.
- Hallucination (AI)
- A hallucination is output from an AI model that sounds fluent and confident but is false, unsupported by its sources or made up, such as an invented citation, quote, statistic or API function.