Zipformer Korean: The Fastest Offline Korean Dictation for Mac
Zipformer Korean is the smallest and fastest model in Resonant. At 68 MB, it's a smaller download than most apps. It runs at a real-time factor of 0.034, which means it transcribes Korean speech 29 times faster than real-time. A full minute of speech processes in about two seconds on Apple Silicon.
Purpose-built for Korean
Korean is a morphologically complex language with a rich agglutinative structure. General-purpose multilingual models tend to underperform dedicated Korean models because the language requires training data and architecture choices that specifically account for Korean phonetics and morphology.
Zipformer Korean is an Icefall Zipformer transducer model trained specifically on Korean speech data. The Zipformer architecture is known for its efficiency — it achieves strong accuracy at a fraction of the parameter count of larger models — and training on dedicated Korean data means it handles the language's specific patterns accurately.
29x real-time is genuinely fast
An RTF of 0.034 is exceptionally fast for a full-quality speech recognition model. To put it in concrete terms: if you dictate a 30-second note, the transcription is complete in about one second. For quick voice-to-text — Slack messages, short memos, form fields — it's essentially instant.
This speed also means the model has very low CPU impact while running. On a MacBook with other applications open, Zipformer Korean won't compete meaningfully for resources.
When to use it
If Korean is your primary or only dictation language, Zipformer Korean is the right choice. The 68 MB download is negligible, the speed is unmatched, and it's trained specifically for Korean.
If you regularly switch between Korean and other East Asian languages (Chinese, Japanese, Cantonese), consider SenseVoice Small instead — it handles all five East Asian languages in a single model and still downloads in under 250 MB.
How to enable it
Open Resonant Settings → Transcription and select “Zipformer Korean”. The 68 MB download completes in seconds on most connections. The model switches immediately and all processing runs locally on your Mac.
Your Korean speech is never transmitted anywhere. It processes on your machine and the text appears where your cursor is.
Download Resonant for offline Korean dictation on your Mac.