Speech-to-text describes the end-to-end task of turning recorded or live speech into text. It is used for meeting transcripts, voice input, call analysis, captions and searchable audio archives. Depending on the system, the output may include timestamps, punctuation and speaker diarization.
A meeting tool might record a call, transcribe it with speaker labels, then pass the transcript to a large language model for a summary and action items. Each stage can add errors, so keep the transcript beside the summary and let readers check a claim against the original text. Live captioning prioritizes speed, while batch transcription can prioritize accuracy.
Accuracy is affected by noise, overlapping speakers, accents and specialist vocabulary. For reliable workflows, keep the original audio where appropriate, measure transcription errors on representative samples and define when a person must review the result. Related: Speech Recognition, Text-to-Speech.