01 / LOCAL AND PRIVATE
Your Mac. Your words.
Your audio and transcripts stay on your Mac. No saved audio or transcript history. Recordings are temporary and deleted after processing.
No subscriptions. No accounts. No cloud services or API keys.
Whisp Beta
Runs locally and works offline, powered by Whisper. Lightweight by design, it barely uses resources while idle and frees memory after each transcription.
Download Whisp for Mac 3.3 MBApple Silicon · macOS 13.3+ · Signed & notarized
01 / LOCAL AND PRIVATE
Your audio and transcripts stay on your Mac. No saved audio or transcript history. Recordings are temporary and deleted after processing.
No subscriptions. No accounts. No cloud services or API keys.
02 / SMALL BY DESIGN
A 3.3 MB download. Just 9 MB on disk (excluding the speech model). Powered by whisper.cpp, bundled as an XCFramework. No Homebrew, Python, or extra tools to install.
The default English model is a separate 148 MB download during setup.
03 / MEMORY, RESPECTED
Whisp loads the speech model when you need it and releases it after transcription. macOS file caching keeps repeat loads fast without keeping the model permanently loaded.
04 / OUT OF YOUR WAY
Lives quietly in your menu bar, using approximately 0.1% CPU at idle. Works across all your apps, even the terminal. If it accepts text, it works.
Idle CPU usage varies by Mac and system activity.
05 / MODELS & LANGUAGES
Bring your own custom model, such as one fine-tuned for your language, or add the multilingual model supporting 99 languages. With a multilingual model, choose Auto-detect Speech to recognize the language you’re speaking, or set a fixed language manually.
06 / AUTOMATIC LANGUAGE SWITCHING
Whisp picks up supported languages from your configured keyboard layouts. Choose a model for each language, then enable Auto-match Keyboard. As you switch layouts, Whisp follows automatically, using the model you chose for each language.
Your voice. Your hardware. Your words.