Stop renting your dictation.
$29 once. Hold a key, talk, let go — it's typed, and cleaned up.
Re-created at real talking speed. Nothing is sped up, and Wi-Fi stays off the whole time.
What you said, and what got pasted:
Raw: um so i was thinking we should uh ship the the new version on friday no wait i mean thursday can you uh send me the numbers before that. Cleaned: So, I was thinking we should ship the new version on Thursday. Can you send me the numbers before that?. This demo is a scripted demo, re-created at real speed.
One take, end to end
any text field
So, I was thinking we should …
Hold a key. Talk. It lands.
Hold ⌥ and speak
- ⌥hold · walkie-talkie
- ⌥tap once · hands-free
- ⌥hold to discard · hands-free
- escalso cancels, while holding to talk
Your Mac cleans it up
A speech model transcribes, then a small local AI:
- the umsumremoved
- punctuationinsteadinstead?
- namesdungDung,
It lands at your cursor
Pastes wherever you were typing.
median1.4 sto clean text · M4 Max
All three steps happen on your Mac. There is no step four where something gets uploaded.
Works wherever you can type.
Myna types where your cursor is. If you can type there, you can talk there — even in Google Docs, whose own voice typing is cloud-only and stops at the browser tab.
- Cursor
- ChatGPT
- Slack
- Google Docs
- Notes
- Obsidian
- your terminal
- anywhere
Instant by default, polished on demand.
Two modes in one app; a single toggle flips between them. Both included in the one purchase.
- Instant — as spoken
Your words exactly as you said them, the moment you let go. Verbatim, no cleanup.
- Polished
Adds local AI cleanup: the ums go, punctuation lands, your names spell right. Reads like you wrote it.
Measured, not promised
Fast enough to feel instant — and exactly this fast with no internet.
There's no server to wait for, so it's exactly this fast on a plane, on café Wi-Fi, or when someone else's servers are having a bad day.
Slower Macs, slower results.
How we measured this
Medians from 556 real dictations on an M4 Max, July 2026, timed from key-release to text at your cursor. Instant is a bare transcript (0.8 s); polished adds the local cleanup (1.4 s). Slower Macs, slower results. The models stay warm in memory, so there is no cold start.
median phrase: 124 characters from 12 s of speech
Own it
$29 once. No account, no subscription, nothing to cancel.
Offline
Runs entirely on your Mac. Works on a plane, on café Wi-Fi, when someone else's servers are down.
Cleans your English
Fillers, stutters, self-corrections, punctuation, names — fixed as you talk.
Just works
One hotkey. Nothing to configure, no API keys to paste.
The small things that make it stick.
- Cursor
- Slack
- any text field
Custom vocabulary
vocab.txt
Đà Nẵng
Nguyên
da nangĐà Nẵng
One text file; both models enforce it. Edits apply on your next take.
One toggle, two modes
0.8 sverbatim
Both included in the one purchase. Flip anytime.
Two ways to talk
⌥hold · walkie-talkie⌥tap once · hands-free⌥hold to discard · hands-freeescalso cancels, while holding to talkHold like a walkie-talkie, or tap for hands-free. Esc cancels.
Live preview
listening…not in your screen shares
Words form as you speak; invisible to screen recordings and shares.
A quiet menu-bar app
Starts at login, stays out of the way.
Paste-again rescue
Edit vocabulary…
Pasted into the wrong window? Re-paste your last dictation from the menu bar — the most recent one only, kept on your Mac and overwritten every time.
If English isn't your first language, this is the part you'll notice.
Say it the way it comes out — false starts, fillers, grammar and all. What lands is the sentence you meant, in clean English.
DevsWritersFoundersESL speakers who want clean English
Where your voice goes, and what it costs.
| Myna | Cloud dictation apps | Built-in macOS | |
|---|---|---|---|
| Your audio | Stays on your Mac | Uploaded to their servers | Mostly on-device |
| Cleanup | Local: fillers cut, punctuation fixed, your vocabulary enforced | Yes, in the cloud | Auto-punctuation only, no cleanup. Fillers land verbatim or get misheard. |
| Cleanup runs where? | On your Mac | On their servers | No cleanup |
| Works offline | Always | No | Partly |
| Account required | No | Yes | No |
| Price | $29 once | $12–15 a month | Free |
| Cost over two years | $29 | $288–360 | $0, but raw |
Your audio
- Myna
- Stays on your Mac
- Cloud dictation apps
- Uploaded to their servers
- Built-in macOS
- Mostly on-device
Cleanup
- Myna
- Local: fillers cut, punctuation fixed, your vocabulary enforced
- Cloud dictation apps
- Yes, in the cloud
- Built-in macOS
- Auto-punctuation only, no cleanup. Fillers land verbatim or get misheard.
Cleanup runs where?
- Myna
- On your Mac
- Cloud dictation apps
- On their servers
- Built-in macOS
- No cleanup
Works offline
- Myna
- Always
- Cloud dictation apps
- No
- Built-in macOS
- Partly
Account required
- Myna
- No
- Cloud dictation apps
- Yes
- Built-in macOS
- No
Price
- Myna
- $29 once
- Cloud dictation apps
- $12–15 a month
- Built-in macOS
- Free
Cost over two years
- Myna
- $29
- Cloud dictation apps
- $288–360
- Built-in macOS
- $0, but raw
Built-in macOS dictation is free; if literal transcription is enough for you, use it.
Priced against the field.
From each maker's own pricing page, July 2026.
- Wispr Flow$15/mo · $12/mo annualCleanup runs in the cloud — always; no offline mode.
- Willow$15/mo · $12/mo annualCloud cleanup; offline fallback drops to a weaker local model.
- Aqua Voice$8/mo billed annuallyCloud cleanup; month-to-month price unpublished.
- Superwhisper$8.49/mo · $249.99 lifetimeLocal + cloud hybrid — models, modes, and prompts to configure.
- VoiceInk$25 once · 1 MacLocal transcription; AI cleanup needs your own API key.
- Built-in macOSFreeRaw transcription only — no cleanup at all.
- Myna$29 onceLocal cleanup included. Offline, no account, no keys.
Don't take the table's word for it.
We read the same sentence, fillers and all, to Apple's built-in dictation and to Myna. Single takes, unedited, July 2026.
“so um can you uh look at why the the login test is flaky and um maybe mock the clock instead”
Built-in macOS dictation typed
“so can you look at what (misheard) the login test is like you (misheard) and maybe remove (misheard) the clock instead?”
Myna typed
“So, can you look at why the login test is flaky and maybe mock the clock instead?”
Pay once. Own it.
Cloud tools bill monthly because their servers cost money every time you speak. Myna runs on your Mac — nothing to bill for. That's why it's $29 once, not $15 a month.
$29one-time purchase
$29
once, today
$0
a month, forever
0
accounts to make
- Lifetime license, one-time payment
- All updates free, nothing recurring
- Both modes included — Instant and Polished
Checkout by Stripe. Your download link and license arrive by email.
Full refund within 30 days — no hoopsQuestions, answered plainly.
Why isn't it a subscription?
Because it runs on your hardware. There are no server costs to pass on, so there's nothing to bill monthly. Pay $29 once, own it, and updates are free.
What do I need?
- Chip
- Apple Silicon, M1 or newer (Intel not supported)
- macOS
- 14 or later
- Disk
- ~3 GB free for the app and its models (2.8 GB fully loaded)
- RAM
- 16 GB recommended. On 8 GB, Instant runs well; Polished needs more memory than the Mac can spare, so it stays off
- Permissions
- Microphone and Accessibility, granted on first launch — the two it needs to hear you and type into any app
- Download
- ~127 MB installer. The 1.2 GB speech model downloads on first launch; the 1.6 GB cleanup model when you first turn on Polished. Then it runs offline, for good.
Does it work in my app?
Anywhere you can place a text cursor: editors, browsers, AI chats, terminals, email, Slack. Myna pastes at the cursor, so it needs no per-app integrations.
Will it spell my names right?
Add them to your vocabulary file once: one term per line, plus optional "what it hears => what you mean" corrections. Both the transcriber and the cleanup model enforce the file, and edits apply on your very next dictation.
Does it make things up?
No. Myna won't hallucinate words into silence or background noise — a common failure of speech models that type phantom phrases like "thanks for watching" into a quiet mic. If you didn't say it, it won't type it.
Is it private?
It runs on your Mac — there's no server, no account, nothing to leak. Not a privacy pitch, just how it's built.
How do I get it after I buy?
You'll get an email with your download link and license — I send them by hand right now, usually within a day, to the address on your receipt. Any trouble, email d@itechnoir.com.
Why isn't it in the Mac App Store?
Pasting into any app needs an accessibility permission the App Store sandbox forbids. Myna ships as a signed, notarized direct download instead; macOS checks the signature on first launch.
I handle confidential material. Can I use this?
No app makes you HIPAA-compliant by itself, and I won't pretend otherwise. What's structurally true: dictation happens entirely on your Mac, so there's no vendor processing your recordings, no cloud transcript, and no third-party to vet. Ask your compliance people, and bring this page.
Your Mac can already do this.
An M-series chip is a very capable dictation machine. Myna is the app that lets it act like one.
Requires an Apple Silicon Mac (M1 or newer), macOS 14 or later. 16 GB RAM recommended, about 5 GB free disk.