Encyclopedia · 176 concepts

NLP & Language · intermediate · concept 89 of 176

Speech Recognition (ASR) & Whisper

Turning audio into text. Whisper made near-human transcription open source and free to run on a laptop, trained on 680,000 hours of web audio in 90-plus languages. ASR is the front door of voice assistants, meeting notes, subtitles, and every voice-first AI interface.

Key terms

ASRMel spectrogramWhisperWord error rateDiarization

Where you meet it in the real world

Meeting transcription, voice assistants, call-center analytics, subtitles, medical dictation

Videos