~$2B
Cloud dictation SaaS. ~$12–15/mo. Free tier caps. Your audio leaves the machine. Marketing is the product story.
Open source · Mac only · $0
Stop renting your voice to a $2B cloud app.
Wispr Flow is valued around two billion dollars. That money buys marketing — ads, stunts, and a story. MacWispr does the same job free, on your Mac, in under half a second. Open source must win.
MIT open source
On-device ASR
No account
No word limits
$0
Local on Apple Silicon. <0.5s. MIT licensed. No account. No meter. Code you can fork if we suck.
The problem
You’re paying a subscription for something your Mac can already do
Cloud dictation adds latency, privacy risk, and another monthly bill — then spends the raise on the narrative that you need them forever.
The tax
~$12–15/mo (or more). Free tiers hit word walls. Talking becomes metered content.
The wait
Upload → queue → polish → paste. ~2–3s feels fine in a demo, slow in real work.
The leak
Your voice leaves the device. Wrong default for code, notes, and anything private.
Why a $2B price tag stings
Marketing is not a moat. Defaults are.
A giant valuation funds ads and launch theater. It doesn’t make a cloud round-trip faster than your Neural Engine — and it doesn’t make open source less free.
They buy attention
Stunts, retargeting, “3× accuracy” slogans. Attention is rented. Habits are too — at $15/mo.
We ship the default
Menu bar. ⌥Space. Text lands where you are. Local model. Updates via Sparkle. No deck required.
Open source must win
Inspect it. Fork it. Leave anytime. The category is infrastructure — like a keyboard, not a lifestyle brand.
Straight comparison
What actually matters when you dictate
We don’t try to win every category — only the ones that change your day.
| What you care about | MacWispr | Wispr Flow |
|---|---|---|
| Price for core dictation | $0 forever | Paid (~$12–15/mo) |
| Where audio runs | Your Mac (local) | Their cloud |
| Typical feel | <0.5s | ~2–3s network tax |
| Account required | No | Yes |
| Word limits | None | Free tier caps |
| Source | MIT open source | Closed |
| Platform | Mac / Apple Silicon | Cross-platform |
What people keep saying
The timeline already voted with cancellations
Paraphrased public sentiment — the recurring themes, not the ads.
“A $2B valuation doesn’t look convincing when the core feature becomes free.”
“Cancelled my paid plan. Free, local, open source on my Mac.”
“Local is faster. Local is private. Free. No marketing spam.”
How we solve it
Same job. Better defaults.
One clear promise: free, local speech-to-text for Mac — without the SaaS tax.
Pain
Subscription fatigue
Another AI bill for talking. Free tiers wall you off mid-week.
Fix
$0. MIT. No meter.
Core STT is free forever. Optional BYOK polish if you want it — never required.
Pain
Cloud lag in real work
Fine in demos. Deadly in Cursor, Slack, Mail when you’re in flow.
Fix
<0.5s on-device
Qwen3-ASR via MLX on Apple Silicon. Your GPU. No region hop.
Pain
Voice leaves the machine
Wrong default for private thoughts, client notes, and code.
Fix
Local by default
Audio stays for local ASR. Telemetry opt-in only — never transcripts. Privacy contract.
Pain
Offline dead zones
Flights and bad Wi‑Fi kill cloud tools when you need them most.
Fix
Works offline
Model lives on the Mac. Airplane mode is a feature, not a ticket.
The bet
David doesn’t need $2B. You need a better default.
We’re not outspending a giant on ads. We’re out-defaulting them: free, local, open source — so open source wins the category that should never have been rented.
macOS 14+ · Apple Silicon · 3-step install · no account