Speech-to-Text
Also called STT, Speech recognition, Automatic speech recognition, ASR, Transcription
AI that converts spoken audio into written text, used for transcription, captions, meeting notes and voice assistants.
What is speech-to-text?
Speech-to-text (STT), also called automatic speech recognition (ASR), turns spoken words in audio or video into text. It powers transcription apps, live captions, meeting note-takers, voice typing and the "listening" half of voice assistants.
How does it work?
The audio is converted into a visual-like representation of its sound frequencies, then a model, usually a transformer, predicts the text that matches it. Modern models are trained on very large amounts of recorded speech in many languages, which makes them much better at accents and background noise than older systems. Some also add punctuation, label who is speaking (diarisation) and attach timestamps.
Speech-to-text can run in the cloud or, with open models such as Whisper, on your own computer.
What is it commonly confused with?
- Text-to-speech. That's the reverse: turning text into spoken audio.
- Understanding. A transcript records words, not meaning. Summaries and action items from a meeting tool come from a separate language model reading the transcript.
Why does it matter?
Transcription makes audio searchable, accessible and easy to summarise. When comparing tools, check accuracy in your languages and accents, speaker labelling, whether it works live or only after recording, and where the audio is stored. Transcripts can still mishear names, numbers and technical terms, so check anything important. Recording people usually requires their consent, and in some places that's a legal requirement.
