Speech engines and models
Choose in Settings → STT → Speech Recognition Engine.
| Engine | Where it runs | Needs | Notes |
|---|---|---|---|
| WhisperKit (local) | On your Mac | A model download | Picking a model in the tutorial selects this engine. Supports live preview. Pause line breaks work only here. |
| MLX Audio (local) | On your Mac | uv installed | The default when you skip the tutorial and uv is available. The default model is Qwen3-ASR-1.7B-8bit. |
| Groq Cloud (fast) | Cloud | Groq API key | Audio is sent to Groq. Test connection under the key checks that it works. |
| Apple (built into macOS) | On your Mac | macOS 26 or later | Uses the speech recognition built into macOS, with no model file to download. The OS downloads the language assets the first time. Supports live preview. |

Choosing between Apple and WhisperKit
Measured on macOS 27 with 120 sentences from a public Korean read-speech dataset (18.4 minutes, 10 speakers) and a 145-second real lecture recording. Figures are character error rates on sentences without numbers.
| Condition | Apple | WhisperKit Large v3 Turbo (default) |
|---|---|---|
| Clean read speech | 3.3% | 3.0% |
| Read speech + babble noise | 6.0% | 6.4% |
| Sentences more than 30% wrong (noise) | 6 of 120 | 10 of 120 |
| Time for 18.4 minutes | 14 seconds | 240 seconds |
- On clean speech WhisperKit is slightly more accurate, and Apple is 17 times faster.
- With noise or on long real recordings, Apple broke down less. WhisperKit Large v3 Turbo made up and repeated sentences in one stretch of the lecture recording. Apple's errors are word-level, such as writing
범용(general-purpose) as법용. - The most accurate is WhisperKit Large v3. It was the most accurate of the three on the same lecture recording, and it is the slowest.
- Apple spells English terms in Hangul (
isolation→아이솔레이션). - Apple does not detect the language. With the transcription language set to
Auto-detect, it listens for Korean.
Choosing a WhisperKit model
A model downloads as soon as you pick it. See progress and sizes in the Settings → Models tab.
| Model | Size | Character |
|---|---|---|
| Tiny | about 77MB | Fastest, but hard to use for Korean. |
| Base | about 147MB | Often gets proper nouns and technical terms wrong. |
| Small | about 486MB | The lower bound where Korean becomes usable. |
| Large v3 Turbo | about 3.2GB | Default. Balanced speed and accuracy |
| Large v3 Turbo (v20240930) | about 1.6GB | A later snapshot of the same family |
| Large v3 Turbo compressed | about 632MB | Smaller and faster than the default, with lower CPU use. |
| Large v3 quantized | about 947MB | A smaller version of Large v3 |
| Large v3 | about 3.1GB | Most accurate and slowest. |
If transcription is slow, check Activity Monitor for other apps using the CPU before switching models. When specific terms keep coming out wrong, adding them to the Dictionary is quicker than a bigger model.
Models are stored in ~/Library/Application Support/Sogon/Models/whisperkit.