You need an Apple Silicon Mac running macOS 14 or later, enough free space for a speech model, and permission to use the microphone. Accessibility permission lets SuprFlow insert the result into the app where you are typing. Start with a harmless test sentence and the default local cleanup setting. You can evaluate optional AI formatting later.
The settings labels below were checked in the installed SuprFlow 1.0.0 (12) app on 28 September 2026. That check covered the settings screens, not the onboarding flow or the app's network behavior. If you are installing for a firm, confirm the signed download and its privacy documentation with your IT owner before entering confidential material.
1. Check the Mac and download source
Open Apple menu → About This Mac and check the chip and macOS version. SuprFlow's requirements page currently lists Apple Silicon and macOS 14+. Get the app from the official download page, then follow the installer and any macOS verification prompts. A model download is required on first setup; keep a reliable connection and enough disk space until the app says the model is ready. Use the requirements page as a planning estimate and check your model choice and actual disk usage.
2. Pick the language and wait for the model
The onboarding flow asks for a language and downloads a speech-recognition model. In the installed app, open General → Speech recognition language to check the language, then Speech → Speech recognition model to check the model. Version 1.0.0 (12) shows Small, Small (English), Medium, and Medium (English) cards; it marks the loaded model Active under Storage. Small is described as about 460 MB and Medium as about 1.5 GB. Those labels are model sizes, not a measured total first-download figure. If setup stalls, check the connection, storage, and whether a managed network blocks the model host. Test the active model with your accent and vocabulary before downloading another one.
3. Grant the two working permissions
Microphone permits recording. Accessibility permits SuprFlow to type into the focused app and enables the Fn-key event tap. The installed app shows both permission statuses under General → Permissions. macOS presents prompts during onboarding. If you skipped them, open System Settings → Privacy & Security → Microphone and Accessibility and enable SuprFlow, then return to the app. On a managed Mac, ask IT if those controls are locked.
The onboarding screen says text can fall back to the clipboard without Accessibility. Treat that as a fallback, not a successful insertion test. If dictation completes but nothing appears in the intended text field, check the Accessibility setting, confirm that the target field still has focus, and try a simple text editor with synthetic text. A browser address bar, password field, secure app, or remote desktop can behave differently from an ordinary text field.
4. Set your hotkey and make one test
Open Hotkey and choose Fn Key to hold Fn while recording and release it to transcribe, or Custom Hotkey to press once to start and again to stop. The screen has a test text field labeled “Your transcribed text will appear here…”. Its help says to press Space while holding Fn to lock hands-free recording. Try the test field first; then try a separate app to confirm insertion. Choose a custom shortcut that does not collide with macOS or another app.
Click into a blank local note and dictate: “Send the sample draft on Thursday. New paragraph. Check the number forty-two.” Release or toggle the hotkey to finish. Confirm the sentence arrives in the note and inspect punctuation and the number. If the Fn key opens the emoji picker, the app's Hotkey screen says Accessibility may be missing or its event tap unavailable. Grant permission and retry; a custom shortcut can also avoid the conflict.
5. Choose a data path before real work
Open Formatting. Version 1.0.0 (12) offers Default (“Essential formatting — punctuation, capitalization, and number cleanup”), Cloud API, and Local (Ollama). Default is the local cleanup path with no AI provider. Cloud API sends transcript text and formatting instructions to the selected provider or custom endpoint. Ollama can be local when its endpoint is a server on this Mac; a configured remote endpoint is a network path. Record the provider and endpoint before dictating work material.
Open General → Enable clipboard to check whether clipboard history is on; its subtitle says it enables clipboard history and the ⌥⌘V shortcut. Open Transcriptions to inspect local history. SuprFlow stores recent transcription entries locally. Clipboard history is separate and may retain other copied content; check the switch on your own Mac. Release builds also include usage telemetry and account/license traffic, which are separate from local speech recognition. See the data-flow guide for the full distinction.
If your first dictation fails
| What happened | First thing to check |
|---|---|
| Hotkey does nothing | Hotkey settings, keyboard conflict, Accessibility permission, and whether the app is running in the menu bar. |
| Recording starts but no words appear | Microphone permission, input device, model-ready status, and selected language. |
| Words appear in history but not the app | Focused field, Accessibility permission, secure-field restrictions, and clipboard fallback. |
| Result sounds correct but names or numbers are wrong | Try a dictionary correction and repeat a short synthetic example before trusting a real document. |
| Formatting is slow or fails | Check whether Cloud API or Ollama is selected and whether that endpoint is reachable; test the local default separately. |
If a transcription completed but did not land in the text field, the installed app's Help screen says Control-Command-V opens a picker of recent transcriptions. Select the intended field before using it and check the result. You can also copy a saved entry from Transcriptions. This recovery shortcut is distinct from the optional clipboard-history shortcut Option-Command-V.
Questions people ask
Do I need to sign in before the local model can transcribe?
The current onboarding offers sign-in while models download, and the app has account/license checks. The exact entitlement and offline grace of a signed release need release-specific verification. Do not promise indefinite use without an account or connection based on the speech model alone.
Can I use SuprFlow without internet?
The speech model must be downloaded first. Local speech recognition and default cleanup can then run on device. Account/license, updates, release telemetry, and any chosen hosted formatter have separate network behavior. An organization should test its approved signed build offline.
Should I turn on hosted AI formatting immediately?
No. First establish whether the local output is good enough for your job. If you later choose a hosted formatter, review the provider's data handling and training terms and use only material your policy allows you to send.
Related guides
- Change the SuprFlow hotkey
- Check Bluetooth microphone quality
- Where does my dictation go?
- Does SuprFlow use dictation for AI training?
- SuprFlow vs Apple Dictation
- Make SuprFlow spell names and acronyms correctly
- Use voice snippets on Mac
- Choose local or cloud formatting
- Fix SuprFlow not pasting into an app
- Fix the Fn key opening emoji instead of recording
- Fix a speech model download that appears stuck