Transcribe
Models & languages
Choose and download a transcription model for your language and your Mac.
In the appSettings → Engine
On this page
Supported languages by engine
Whisper supports 99 languages, Nemotron 80, and Parakeet v3 25. Apple Speech availability varies by Mac. Visit Languages & automatic detection for the full reference, Parakeet’s language list, and how each engine uses your spoken-language choice.
Set up local transcription
- Open Settings → Engine → Transcription engine.
- Choose Whisper, Parakeet, or Nemotron.
- Select a row in the Models card.
- Press its Get control and wait for downloading and preparation to finish.
Selecting and downloading are separate actions. Ready means a model is loaded for transcription. The default model within the local catalog is Parakeet v3 (quantized).
After the initial download, the model transcribes on your Mac without sending audio to a provider.
Choose a local model
Choose Parakeet for fast transcription in its 25 languages, Whisper for coverage across 99 languages, or Nemotron to see words while you speak in any of its 80 languages. Compare a short example using your own voice, microphone, and specialist words. Visit Usage to compare processing times and activity for models you have used.
| Engine | Model | Download, approximately | What to know |
|---|---|---|---|
| Parakeet | Parakeet v3 (quantized) | 637 MB | Default local choice; English and 24 European languages. |
| Whisper | Large v3 Turbo (quantized) | 574 MB | Whisper with broad language coverage: 99 languages. |
| Whisper | Base | 148 MB | A smaller, quick-to-download Whisper model. |
| Whisper | Tiny | 78 MB | The smallest download in this catalog. |
| Nemotron | Live | 672 MB | Live transcription in 80 languages, showing words about a second behind your speech. |
| Nemotron | Live, steadier | 672 MB | The same 80 languages, with words arriving a little later for steadier transcription. Recommended for Chinese and Japanese. |
| Parakeet | Parakeet v3 | 1,197 MB | The full-precision Parakeet option. |
| Whisper | Large v3 Turbo | 1,620 MB | The full-precision Whisper Turbo option. |
| Whisper | Medium | 1,530 MB | A larger multilingual Whisper option. |
| Whisper | Small | 488 MB | A middle-sized Whisper option. |
| Whisper | Small (English) | 488 MB | English only. |
| Whisper | Base (English) | 148 MB | English only. |
| Whisper | Tiny (English) | 78 MB | English only. |
Download sizes are approximate. Quantized models use a more compact representation to reduce storage and memory needs. Memory use while running varies by model.
Manage downloads
The Models card shows downloads for your selected engine. Choose Show all models to expand beyond the recommended choices, then filter by name, size, or language. Cancel a download with its cancel control. Remove an installed model using its trash button.
If a model repeatedly fails to load, delete it and download it again. Allow preparation to finish before trying to use it. Deleting a selected model means you need to download it again or select another installed model.
Choose the language you speak
Use Language → Spoken language with Whisper, Nemotron, Apple Speech, or a cloud provider. The picker offers Automatic plus 29 named languages. Parakeet always detects the language itself.
Automatic detects the language with Whisper and Nemotron, but uses your Mac’s current language and region with Apple Speech. See what Automatic means for each engine.
To write in a different language, use a translation rewrite profile. Spoken language describes the audio you provide.
Parakeet v3: supported languages
Parakeet detects 25 European languages automatically and accepts no language setting. MicMyDay displays an explanation in place of the language picker, with a warning if your saved choice is not supported by Parakeet.
Use Apple Speech
Choose Apple Speech in the engine picker and allow Speech Recognition permission. Its Automatic setting uses your Mac’s current language and region, so select a language explicitly when you speak another one.
Language availability depends on this Mac’s language resources. Read about Apple Speech availability and on-device recognition.
Apple Speech transcribes audio. Apple built-in under Rewrite is a separate feature for editing text.