Transcription and on-device AI
How speech recognition on a Mac works, how accurate it is, and how to get a better transcript.
Start here
More in this topic
- Transcription and on-device AIApple's SpeechAnalyzer: speech recognition in macOS 26What Apple shipped for speech-to-text in macOS 26: SpeechAnalyzer and its transcribers, their languages, and how they differ from SFSpeechRecognizer.
- Transcription and on-device AICan you trust an AI meeting summary?Where AI meeting summaries go wrong — invented commitments, missing dissent, wrong names — and a quick way to check one against the transcript before you send.
- Transcription and on-device AIFinding a moment in a meeting recordingHow to find a moment in a meeting recording without listening to all of it - search the transcript, jump from a timestamp, or use an overview of who spoke when.
- Transcription and on-device AIGetting better meeting transcripts: microphones, rooms and habitsPractical ways to get more accurate meeting transcripts on a Mac - headphones, the right microphone input, distance, one voice at a time, names spelled out.
- Transcription and on-device AIHow to transcribe a meeting offline on a MacTranscribe a meeting on a Mac with no internet connection: what to download first, what works offline and what does not, and how to check nothing is uploaded.
- Transcription and on-device AILive transcription vs transcribing after the meetingLive transcripts are fast enough to follow a meeting; a pass over the whole recording afterwards can be more accurate. Why, and what it takes to have both.
- Transcription and on-device AINaming the speakers in a meeting transcriptWhy transcripts say Speaker 1 and Speaker 2, how to label speakers in a transcript with real names, and how to fix lines credited to the wrong person.
- Transcription and on-device AIRecognising voices across meetings: the privacy questionsA notetaker that recognises voices across meetings stores voiceprints, which GDPR and Illinois BIPA treat as biometric data. What to ask before switching it on.
- Transcription and on-device AISpeaker diarization: how transcripts work out who spokeSpeaker diarization splits a recording by who spoke when. How it works, how it differs from recognising voices, and what two tracks give you for free.
- Transcription and on-device AIWhich languages can a Mac transcribe on the device?The ten languages Apple's on-device SpeechTranscriber supports on macOS 26, with regional variants, how dictation differs, and options for other languages.
- Transcription and on-device AIWhy meeting transcripts get names and numbers wrongNames, numbers and jargon are the hard words for every speech recogniser. Why, which errors matter, and what the room and microphone change.
- Transcription and on-device AIWhy Notey re-transcribes a meeting after it endsAfter an English meeting, Notey reads the recording again on your Mac and keeps the words the recognisers agree on. When it runs and how to turn it off.
- Transcription and on-device AIWord error rate, explained for people choosing a transcription toolWhat word error rate counts, why vendors' accuracy figures are hard to compare, and how to test a transcription tool on your own meetings.