Apple's speech-to-text APIs
The transcription APIs built into macOS - useful, free, and not built for multi-speaker meeting audio.
Apple ships speech recognition with the operating system. SFSpeechRecognizer has been there for years and is what dictation-style features use; newer macOS releases added a more modern on-device transcription API for longer audio. Both are free, need no download, and are a sensible choice for single-speaker dictation.
They are a weaker fit for meetings. They transcribe one stream of speech without telling you who spoke, so you still need a separate diarization step and a way to keep names stable across recordings. Control over the model is limited, and behaviour varies by macOS version, which is awkward for an app that has to produce the same transcript on every supported Mac.
Mac Note Taker uses Parakeet models through Core ML instead, so the same transcription quality, speaker separation, and language coverage are available on every supported version of macOS.
Related terms
- ASR (Automatic Speech Recognition) ↗The model layer that converts spoken audio into written text. The first stage of any meeting transcription pipeline.
- Core ML ↗Apple's on-device model runtime. It decides whether a model runs on the Neural Engine, the GPU, or the CPU.
- Whisper ↗OpenAI's open-source ASR model. Whisper Large v3 is Mac Note Taker's high-accuracy multilingual option.
- Parakeet TDT ↗NVIDIA's open-source ASR model family. Parakeet TDT v3 is Mac Note Taker's default English transcription model on the Neural Engine.