ASR (Automatic Speech Recognition)
The model layer that converts spoken audio into written text. The first stage of any meeting transcription pipeline.
ASR is the umbrella term for the model class that turns an audio waveform into a sequence of words. In a meeting transcription stack, ASR is the first stage after audio capture and voice activity detection. The most common open-source ASR families in 2026 are Whisper (OpenAI), Parakeet (NVIDIA), and Distil-Whisper.
On an Apple Silicon Mac, modern ASR models are converted to CoreML and executed on the Apple Neural Engine, which makes real-time transcription cheap enough to leave running for the duration of a meeting. Mac Note Taker uses Parakeet TDT v3 by default and can swap to Whisper Large v3 when higher multilingual accuracy is needed.
Related terms
- 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.
- Whisper ↗OpenAI's open-source ASR model. Whisper Large v3 is Mac Note Taker's high-accuracy multilingual option.
- VAD (Voice Activity Detection) ↗A lightweight model that flags which slices of audio contain human speech, used to gate the heavier ASR and diarization stages.
- ANE (Apple Neural Engine) ↗Shorter name for the Apple Neural Engine - the per-chip ML accelerator that runs Mac Note Taker's transcription and diarization models.