SWIFT · WHISPER · ON-DEVICE AI · Cork AI Consulting
BlarneyKeyBlarneyKey — Private, Offline AI Dictation for macOS
Cloud dictation sends every word you say to someone else’s server, and charges monthly for it. Douglas and I built BlarneyKey to prove the opposite: fast, accurate speech-to-text that runs entirely on the Mac, free, and open on GitHub.
- 0 bytes of audio sent over the network
- €0 free, no subscription
- Open source public on GitHub
- Value €0 Cost to the user
- Time Push-to-talk Hold, speak, release
- Effort Co-built With Douglas Woollam
- Quality 100% On-device
- Step 01 On-device by design
- Step 02 One gesture
- Step 03 Open and free
- Output €0 Cost to the user
01 — Context
The problem.
Dictation is one of the most useful everyday AI tools, and most options send your voice to the cloud and bill you monthly — a poor fit for anyone handling client or personal information.
02 — Method
The approach.
The build, in the order it happened.
-
On-device by design
Whisper runs locally on the Apple Neural Engine, so speed comes from the hardware and privacy is a property of the architecture, not a policy.
-
One gesture
Hold a key, speak, release: the transcript is typed into the active app, with no window to manage.
-
Open and free
Published publicly on GitHub so anyone can read exactly what it does.
- Sources SwiftSwiftUIWhisper
- Step 01 On-device by design
- Step 02 One gesture
- Step 03 Open and free
- Output €0 Cost to the user
03 — A run
What one run looks like.
The steps above, played as the agent would walk them. A reading of the write-up, not a recording of a real run.
run blarneykey
[plan] 3 steps
[step 01] On-device by design
↳ Whisper runs locally on the Apple Neural Engine, so speed comes from th…
[step 02] One gesture
↳ Hold a key, speak, release
[step 03] Open and free
↳ Published publicly on GitHub so anyone can read exactly what it does
[tools] Swift · SwiftUI · Whisper · Apple Neural Engine · macOS
[handoff] Shipped as a free macOS app with a public repository
[exit 0]
04 — Outcome
What changed.
- Shipped as a free macOS app with a public repository.
- Featured as a live demo in the Cork AI Consulting products showcase.
05 — Screens
What it looks like.
Generated from the write-up rather than captured from a client's screen — the shape of the thing, not a leaked one.
Push-to-talk dictation for the Mac that runs Whisper on the Neural Engine — no network, no subscription, nothing leaves the machine.
One gesture, end to end, with nothing leaving the machine.
€0 to use, 0 bytes of audio sent, 100% on-device.
06 — Stack & role
Built with.
- Swift
- SwiftUI
- Whisper
- Apple Neural Engine
- macOS
- Role
- Co-builder, with Douglas Woollam
- Duration
- 2026, shipped
- Client
- Cork AI Consulting
- Period
- 2026