TapTalk
Private by default
Best if you want the fastest path to local dictation on Mac, without accounts or cloud processing.
Faster than your keyboard.
Safer than the cloud.
TapTalk dictates at 150 words a minute into any Mac app. It runs entirely on your machine — no server, no account, no uploads.
Mail · New message
Works in every Mac app
Mail
The math
You think in sentences. You're still typing letters.
Average typing: 40 words a minute. Average speaking: about 150. Same thought, four times the throughput — and you don't have to look at the keyboard.
Typing
40wpm
Average across knowledge workers. Drops further on mobile, in another language, or when you're tired.
TapTalk
~150wpm
Conversational speech. Auto-formatted, fillers stripped, ready to send. About 3.7× the words for the same minute.
A 30-minute typing session becomes 8 minutes of talking. Multiply by every email, every prompt, every Slack thread.
Privacy
Your voice. Stays yours.
Cloud dictation records you, ships the audio to a vendor, transcribes it there, and ships the text back. For a medical note, a legal draft, an investor memo — that round-trip is the whole problem.
TapTalk doesn't have a backend. Whisper runs on your CPU and Neural Engine. No telemetry, no account, no analytics. Airplane mode: still works.
How it works
Three seconds. Three moves.
01 · Hold
Hold Fn from any app.
Cursor stays where it is. A pill appears at the bottom of your screen.
Fn
02 · Speak
Talk like a person.
Filler words, ums, and false starts get cleaned up on the way out.
03 · Release
It types itself.
Text at the cursor — no paste, no window switch. Then the pill fades.
Moving Thursday's meeting to 3 PM.
Features
Small app. Real tooling.
Dictionary
Teach it your jargon.
Languages
Twelve, auto-detected.
HUD pill
95 × 20 pixels of interface.
Shortcuts
Fn. ⌥. ⇪. Your pick.
Inside the app
Three views. One settings panel.
Home
Today
Pull staging logs from yesterday evening — anything between 8 and 10pm with a 500.
14:02
Patient tolerated infusion well, no adverse reactions, next review in two weeks.
11:48
Schreib Lena, dass das Meeting auf Donnerstag 15 Uhr verschoben wird — gleicher Link.
09:17
Yesterday
Rewrite this paragraph so it sounds less corporate and more like how a human would actually say it.
19:31
Honest comparison
Where each tool wins.
Best for
Same job. Very different tradeoffs.
All three can turn speech into text. The real differences are privacy, friction, and how much control you want over the workflow.
Broadest platform reach
Best if you want a polished cross-device product and you are fine with a cloud-first setup.
Most configurable
Best if you want offline options plus a more power-user, tweak-heavy workflow.
Category
TapTalk
Wispr Flow
Superwhisper
On-device processing
✓
✕ cloud
✓ local + cloud models
Works offline
✓
✕
✓
Free tier
1,000 words / day
2,000 words / week desktop · 1,000 iOS
Free tier forever + 15 min Pro trial
Pro price
$15 / mo
$8.49 / mo
Account required
✓ no account
✕ sign-up
✓ license only
Custom dictionary
✓
✓
✓ Pro
Languages
100+
100+
100+
Platforms
macOS
macOS, Win, iOS, Android
macOS, Win, iOS
Feel
Minimal, opinionated
Polished, cloud-first
Power-user, configurable
Checked April 22, 2026 against each vendor's public pricing and feature pages.
FAQ
The short version.
Does my audio ever get sent to a server?
No. Transcription happens on your Mac via Whisper. There is no TapTalk backend. Audio, transcripts, and history live entirely on your machine.
How accurate is local transcription?
On Apple Silicon with the medium model, accuracy is on par with cloud providers for clear speech. The custom dictionary closes most of the remaining gap on proper nouns and jargon.
What does it cost?
Free for 1,000 words a day, forever. Pro is €4.99/mo (€2.99/mo early bird for the first 1,000 users) and removes the limit. No credit card to start. No "contact sales" gauntlet.
Will it work on my M1 MacBook Air?
Yes. Minimum: macOS 13, Apple Silicon or Intel. The default model uses ~150 MB of RAM during transcription.
What languages are supported?
Twelve out of the box: English, German, Spanish, French, Italian, Portuguese, Dutch, Polish, Japanese, Mandarin, Korean, Russian. Auto-detect on by default.
Windows, Linux, iOS?
Not yet. We're shipping macOS first, properly. The transcription engine is portable, so the door is open.
Stop typing. Start talking.
Free for 1,000 words a day. Nothing leaves the machine. No card, no account.