Sogon user guide
Sogon is a macOS dictation and meeting-recording app that lives in the menu bar. It turns your speech into text, optionally polishes it with an LLM, and puts it where your cursor was before you started speaking.
What you can do
| Task | Press | Result |
|---|---|---|
| Put speech at the cursor | Hold ⌃⇧R, speak, release | Transcribed, corrected text lands at your original cursor |
| Polish text you typed | ⌃⇧T | Text in the input window is corrected and inserted |
| Translate a selection | ⌘C twice | The original and the translation appear in a window |
| Act on a selection | ⌃⇧C | A grid of correct, structure, translate, convert, Notes, and search actions |
| Insert a saved phrase | ⌃⇧V | Pick a snippet and insert it at the cursor |
| Ask while showing your screen | ⌃⇧S | The captured image and your words are pasted into the target app |
| Record a meeting | Menu bar icon → Session recording | Record, then transcribe and create minutes or a verbatim transcript |
Translating with ⌘C twice and the ⌃⇧S screen region are off by default. Each guide explains how to turn them on.
Requirements
- An Apple Silicon (arm64) Mac on macOS 14 or later
- Microphone and Accessibility permissions
- For correction and translation, one LLM provider (Amazon Bedrock, OpenAI, Gemini, Anthropic, Groq, or local MLX)
Speech recognition runs on your Mac by default. Without an LLM, Sogon inserts the transcription as is.
New here? Read in this order
Getting from install to your first dictation takes about 10 minutes. The longest step is downloading the speech model (about 1.6GB).
What leaves your Mac
With default settings, audio and transcriptions stay on your Mac. If you choose a cloud LLM or turn on AWS Transcribe, Sogon sends text or audio to that service. You'll find the full list of conditions in Data and privacy.