Glossary
ASR is the technical term that covers every system that turns spoken audio into text — from your phone's voice typing to commercial transcription engines like OpenAI Whisper. It's a 70-year-old research field that became commercially viable around 2014 and reached human-parity accuracy on clean audio in 2022.
How modern ASR actually works
Modern ASR is a transformer neural network. It takes an audio signal, converts it to a mel spectrogram (a visual representation of frequencies over time), then feeds that into a model that outputs a sequence of token probabilities. A decoder (typically beam search) turns those probabilities into the most-likely word sequence. OpenAI's Whisper family are the current standard — trained on 680,000 hours of multilingual audio.
Where ASR fits in your workflow
If you've used Siri, Alexa, Google Voice Typing, YouTube auto-captions, or any AI transcription tool, you've used ASR. The same technology runs in real time (live captions, dictation) and offline batch (this site, Otter, Rev). Real-time ASR is harder because you can't see the future of the sentence; offline batch is more accurate because it can revise.
Accuracy benchmarks in 2026
Modern ASR achieves word error rates of 3-7% on clean, single-speaker English. Drop to 88-94% accuracy on accented English or noisy audio, and 80-88% on multi-speaker overlap or technical jargon. We use Whisper-3 directly without proprietary fine-tuning — keeps general-purpose accuracy strong across languages.
In practice
Drop a 30-minute podcast episode in our uploader. The audio gets converted to a mel spectrogram, fed through Whisper, decoded with beam search, and you get a clean text transcript in 60-90 seconds. Total compute on our side: about 2 seconds of GPU time per minute of audio.
Related terms
Further reading
Try AI transcription
Get started — freeLast updated: September 20, 2026
Frequently Asked Questions
What does ASR mean in transcription?
Automatic Speech Recognition. The umbrella technical term for any system that converts spoken audio into written text.
Why does ASR matter when choosing a transcription workflow?
ASR affects how you understand transcript quality, timing, compatibility, or the technology behind speech-to-text results.
Where can I apply ASR on Transcript.you?
Use the related workflow at /ai-transcription when you want to see how this glossary concept connects to an actual transcription task.
Browse the full transcription glossary or read the complete guide to AI transcription.