Whisper for Mac
Whisper for Mac, without the terminal
OpenAI’s Whisper is a strong speech-recognition model. Getting it into your daily typing on a Mac usually means the terminal. Sayframe is Whisper wired into a system-wide dictation shortcut instead, on-device and with no setup.
What Whisper actually is
Whisper is an open-source speech recognition model released by OpenAI. It is good at turning audio into text, across many languages, and it is free to run yourself. But it is a model, not an app. Using it means picking an implementation, a model size, and a way to feed it audio.
The DIY route
The common path is whisper.cpp, a fast, dependency-light port of Whisper that runs entirely on your Mac. Clone the repo, download a model file (sizes trade speed for accuracy), and run it from the terminal against an audio file. That works well for batch jobs: transcribe a folder of recordings, a podcast episode, a lecture. It is not built for typing. It has no system-wide shortcut and does not place text in the app you are using. Getting live speech into a text field means wiring that yourself.
Sayframe: Whisper wired into how you type
- 1Press your Sayframe shortcut in Mail, Notes, Slack, the terminal, or any Mac app with a cursor.
- 2Speak. Sayframe transcribes on-device with whisper.cpp, with Apple Speech as a fallback engine.
- 3Clean text lands at the cursor. No terminal, no manual model wrangling.
- 4Sayframe manages model sizes and updates for you, balancing speed and accuracy.
If you just want to batch-transcribe files
If your actual need is transcribing a folder of recordings or exporting subtitles for a video project, a CLI tool or a dedicated file-transcription app is a better fit than Sayframe. Sayframe is built for dictation while you work, with speech flowing into the app you are using. It does not focus on batch file processing.