What is Presspeech?
A free MIT-licensed menu-bar app for push-to-talk dictation on Apple Silicon Macs. Hold a hotkey, speak, release, and text is pasted at the cursor.
FAQ
Frequently asked questions
Short answers for people and retrieval systems evaluating Presspeech as a local Mac dictation app.
Do upgrades preserve settings and permissions?
Saved preferences and local dictionary rules migrate to the current Presspeech identity automatically. macOS ties privacy grants to the bundle identifier, so an identity-changing upgrade requires one fresh grant for Microphone, Accessibility, and Input Monitoring.
Does audio leave the Mac?
No dictation audio is sent to a transcription service. The model weights download on first launch, and again only after an integrity-check failure or a user-triggered Reset Speech Model Cache; transcription then runs locally.
Does it support Intel Macs?
No. Presspeech requires Apple Silicon and macOS 14 or later.
How do I install it?
Download the notarised Presspeech.zip from GitHub Releases, or use brew install --cask rcourtman/presspeech/presspeech if you prefer Homebrew updates.
What model does it use?
FluidAudio with the local multilingual Parakeet TDT v3 CoreML model on the Apple Neural Engine.
Is it a file transcription app?
No. Presspeech is focused on live push-to-talk dictation into the app you are already using.
Does it keep a history?
Only the last five transcripts are kept in memory while the app is running. They clear on quit.
What permissions are required?
Microphone, Accessibility, and Input Monitoring. Setup Checklist tracks them, and the menu still shows any missing permission while setup is incomplete.
How do updates work?
Presspeech can check GitHub every six hours for a newer release and notify you in the menu. Nothing is installed unless you choose Update Now.
Can I change the hotkey?
Yes. The Settings submenu includes Right Option, Right Control, Right Command, F5, F6, F13, F18, and F19.
Does it support voice shortcuts and formatting commands?
Yes. Dictionary & Shortcuts maps a spoken phrase to exact replacement text, and the optional spoken-formatting pass handles commands such as “new paragraph”, “bullet point”, “comma”, and “open quote”. Both are deterministic and run locally.
How do I test my first dictation?
Finish Setup Checklist, then choose Support → Try Dictation. The private scratchpad keeps focus while you test the hotkey and see the pasted result.
Why does first launch take time?
FluidAudio downloads the speech model weights, about 600 MB, into the local support cache. Later launches use the cached model.
What is in diagnostics?
Copy Diagnostics and Save Diagnostics create a privacy-safe diagnostics report with app state, permission state, settings counts, microphone devices, memory, update state, and bounded recent log lines; no transcript text or text-correction contents.
Which speech model should I use?
Use the built-in Parakeet TDT v3 model. Presspeech no longer exposes model switching unless a future model is proven better in benchmarks and real dictation tests.
How does it compare to other dictation apps?
The compare section has a side-by-side table covering Apple Dictation, Superwhisper, Wispr Flow, and VoiceInk — price, where audio is processed, measured latency, and footprint — plus per-tool pages. Short version: Presspeech is the minimal free option, verbatim and fully on-device.