ModelRefs / ASR (Automatic Speech Recognition) — AI Glossary
ASR (Automatic Speech Recognition) — AI Glossary
AI models that transcribe spoken audio to text. Modern ASR systems (Whisper, Deepgram, AssemblyAI) use transformer-based encoder-decoder or CTC architectures.
Overview
Modern ASR systems (Whisper, Deepgram, AssemblyAI) use transformer-based encoder-decoder or CTC architectures. Word error rate (WER) is the standard metric. Speaker diarization, language identification, and real-time streaming are common advanced features.
Reference details
| Topic | multimodal |
|---|---|
| Also known as | speech-to-text, speech recognition, STT |
| Last reviewed | 2026-06-24 |
Related terms
Example: Where word error rate misleads
Two systems both report 8% WER. One misses filler words; the other mistranscribes names and numbers. For a meeting summary the first is fine; for a medical or financial transcript the second is unusable. WER weights every word equally, and the words that matter to you rarely are.
Commonly confused with
Transcription is not diarisation. ASR turns audio into words; diarisation attributes those words to speakers. A transcript without diarisation is a wall of text, and many systems that advertise transcription treat speaker labelling as a separate, weaker feature.
When to use it
Reach for it when:
- Audio must become searchable, summarisable text
- Accessibility requirements call for captions or transcripts
- Downstream analysis operates on text
Reach for something else when:
- Evaluating on clean read speech when your audio is noisy and multi-speaker
- Domain vocabulary is specialised and no custom vocabulary is available
- The names and figures matter and nothing verifies them
Referenced by
This term is used by the following ModelRefs references:
Continue your research
Use these connected ModelRefs sections to compare alternatives, inspect implementation paths, and review the evidence and governance boundaries relevant to ASR (Automatic Speech Recognition) — AI Glossary.
Frequently asked questions
What is ASR (Automatic Speech Recognition)?
AI models that transcribe spoken audio to text.
Is ASR (Automatic Speech Recognition) the same as speech-to-text?
Yes — speech-to-text, speech recognition, STT are common aliases for ASR (Automatic Speech Recognition).
What concepts are related to ASR (Automatic Speech Recognition)?
Closely related concepts include tts, multimodal.