Speech recognition

In the Ultra Transcenders books

AI-901

Each book explains Speech recognition in context, with comparison tables and the common traps.

Related terms

• Captioning

  Turning speech in live or recorded video into timed text shown on screen; a speech to text use case that can also filter profanity and display partial results.

• Fast transcription

  Converts recorded audio files to text in one synchronous REST call that finishes quicker than the audio's duration and has predictable latency; part of Speech to text.

• Intent recognition (Speech)

  Retired on 30 September 2025, this Speech SDK feature mapped spoken utterances to intents; use speech to text and then CLU or an Azure OpenAI model instead.

• Language identification

  Speech capability that works out which of a set of possible languages is being spoken: up to 4 when detecting at the start of audio, or up to 10 when detecting continuously. It runs with speech to text or speech translation.

• Real-time transcription

  Speech to text mode that converts streamed audio, from a microphone or file, into text as it is recognised, with interim results along the way. Typical uses are voice input and live captioning.

• Speaker recognition

  Capability of Azure Speech that verifies or identifies a speaker based on their voice. It is a Limited Access feature and differs from speech recognition, which transcribes the words spoken.

• Speech captioning

  Generating captions with timestamps, as SRT or WebVTT, from live or recorded audio using speech to text; profanity can be filtered.

• Speech profanity filter

  Setting in speech to text that decides whether profanity in captions and transcripts is masked (the default), removed or left as is (raw).