Concept
aka STT
aka ASR
Speech-to-Text
Models that convert spoken audio into written text — the inverse of TTS.
Definition
Speech-to-text (a.k.a. ASR) models — Whisper, Conformer, Parakeet — take audio waveforms and emit transcripts. They are essential for accessibility, voice assistants, meeting transcription and indexing audio content. Word error rate is the headline metric.
Common use cases
- Transcription
- Voice agents
- Captioning