1. Home
  2. Glossary
  3. Speech recognition
AI glossary · Media & modalities

Speech recognition

Speech recognition: Speech recognition, also called speech-to-text or automatic speech recognition (ASR), is AI that converts spoken audio into written text. It powers dictation, meeting transcription, voicemail transcripts, and voice commands.

Modern speech recognition uses deep learning models trained on huge amounts of transcribed audio, which is why accuracy jumped in recent years. Tools like Otter.ai, Fireflies, and Zoom's AI Companion transcribe meetings live; phone keyboards and Microsoft 365 offer dictation; OpenAI's Whisper is a widely used open model that developers build on.

Accuracy is high on clear, single-speaker audio and drops with crosstalk, accents the model saw less of, background noise, and specialized vocabulary. Product names, acronyms, and people's names are the most common errors. Many tools let you add a custom vocabulary, which is worth the five minutes.

The bigger issue is consent and confidentiality. Recording and transcribing calls is regulated differently by state, and many organizations require notice to all participants. Transcripts can contain client information, health details, or HR matters, so check where the audio is stored and whether the vendor uses it for training.

Example at work

A consultant runs a transcription tool on every client discovery call, then pastes the transcript into an assistant to extract requirements, open questions, and action items. She corrects two misheard product names, confirms with the client that recording is fine before each call, and stores transcripts in the firm's approved workspace.

Why it matters

Speech recognition turns conversations, the richest and least-captured part of work, into text you can search, summarize, and act on. Used with permission, it eliminates note-taking during meetings and means nothing agreed on a call gets forgotten.

Related terms