One key. Three behaviors.
Hold for push-to-talk. Quick-tap to lock-on, tap again to stop. The mode you want is the gesture you used. Smart, hold, or toggle — your call.
Now in beta for Windows
The fastest way to type is to stop typing.
Press. Speak. Done.
Voice dictation for Windows that disappears into your keyboard. Hold a key, speak, release — your words land wherever your cursor is. No tabs, no waiting, no subscription.
⌨ press any key — Murmur reads scancodes too.
Bring your own keys for
G Groq whisper-large-v3
✦ Gemini 2.5 flash
· Local Whisper soon
What makes it different
Built for the way you actually work.
Capture any key.
Even unnamed media keys. Murmur falls back to scan codes so the Copilot key, your mouse thumb button, or that weird F23 all work.
Your keys. Your data.
Audio goes from your mic to the provider you picked, with the API key you own. Nothing in between.
Pro · v1.1 Catches what the model missed.
A second-pass review reads your transcript like a careful editor — fixing homophones, restoring dropped negations, and rescuing garbled names. Voice and structure stay yours. Set it to never, sometimes, or always.
- Real audio meter — not a fake animation
- Local recents, never synced
- System-tray icon, optional window
- Light + dark, follows your OS
- Voice commands & dictionary
- Chrome extension included
How it works
Three steps. One key.
-
01
Pick a key — any key.
Open Settings → Hotkeys → Capture, then press what you want. Murmur learns its name and its scan code, so even keys Windows doesn't formally know still bind.
Capture hotkeyWinShiftF23 Listening… -
02
Hold it. Speak.
Audio streams to your provider the moment you press. Release and it stops — there's no minimum recording length, no awkward silence detector.
Live -
03
Your words appear.
Wherever your cursor is. Email, terminal, IDE, browser, Discord — it's just typed text. No window switch, no paste step.
your editor
Pricing
Free forever. Pay only if you want more.
What this actually costs you
Drag to your daily dictation.
With Murmur · Groq
$0.18 / mo
With Wispr Flow
$144 / yr
Math: Groq's whisper-large-v3 bills at $0.04 per audio hour. Wispr Flow charges a flat
$12/month. Minutes-per-day × 22 working days × $0.04/60 ≈ what Groq would bill you. Your usage will
not be exactly this; nobody's is.
Free
For everyone, forever.
$0
/ month
- Unlimited dictation, BYO Groq or Gemini key
- Smart formatting basics — voice commands, auto-capitalization, custom dictionary
- Floating dictation pill with live waveform
- Any key as a trigger, even unnamed media keys
- Local history, system tray, dark and light
Pro
For people who type all day.
$8
/ month
Join the waitlist
- Everything in Free
- Accuracy review — second-pass review fixes mishears, dropped negations, and garbled names. Automatic on long recordings.
- AI rewrite styles — formal, casual, concise, email
- Encrypted cloud sync across devices
- Priority transcription routing
Team
For shared deployments.
$14
/ seat / month
Join the waitlist
- Everything in Pro
- SSO via Okta, Azure AD, Google
- Centralized billing and key vault
- Admin policies and audit logs
- Dedicated onboarding
The app
A window worth opening.
State orb, real audio meter, segmented mode picker, recents, tray. Quiet when you don't need it; obvious when you do.
Murmur
Listening
Hold F23 · release to send
"Add bullet points to the deck before the 3pm review."
just now
"Reply: thanks for the heads up — I'll have a draft Friday."
2m
"Push to main and ping the channel when CI finishes."
9m
Murmur in the wild
Anywhere a cursor blinks.
Early users
Quiet praise.
It doesn't keep nudging me. The history pane is enough.
Jonas Pell
Writer · switched from Wispr
I bound it to my mouse thumb button on day one. A week later I noticed I was writing commit messages without lifting my hands off the keyboard. That's the whole pitch.
Elena Marquez
Senior engineer, infra
The Smart mode is the right default. Tap to toggle for long thoughts, hold for quick replies. I haven't touched the settings since the second day, which is the highest compliment I give software.
Reza Akhtar
Product designer
Different priorities
Wispr Flow is the polished, well-funded answer. Murmur is the answer you can read.
They optimize for
- One platform, beautifully — macOS first, the rest later.
- A single subscription that rolls up the model bill into one number.
- A polished cloud history that follows you to every device.
We optimize for
- Windows — including the ARM laptops nobody else has shipped to.
- Letting you keep paying for the model you already pay for.
- A history file you can grep, delete, and walk away from.
The line-by-line table, if you want it.
Murmur
Wispr Flow
Platform
Windows · ARM64 + x64
macOS only
Pricing
Free · BYO key
Subscription
Source
Open · MIT
Closed
Trigger key
Any key — even unnamed
Fixed shortcut
Hotkeys
Two at once: tap + hold
One
Storage
Local history only
Vendor cloud
Reflects publicly available Wispr Flow info as of April 2026. Their product changes; double-check before deciding.
Ready when you are
Stop typing. Start talking.
Download Murmur, drop in a Groq or Gemini key, pick a hotkey. You'll be dictating in under a minute.
Windows 10 / 11 · Intel/AMD64 or ARM64 · A microphone · A Groq or Gemini key
FAQ
Quick answers.
Do I need a paid API key?
Groq's free tier is enough for light daily use. Gemini also has a generous free tier. You'll only need to pay if you dictate hundreds of times a day.
Where does my audio go?
Straight from your mic to whichever provider you set up. Murmur has no server. Audio is streamed, not stored.
Will it intercept the Copilot key?
Yes — but Windows tries first. Disable the stock launcher in Settings → Personalization → Text input, or remap with PowerToys, and Murmur takes over cleanly.
Does it work on ARM64?
Yes — there's a dedicated ARM64 build with a bundled native Python runtime. Pick ARM64 on the download chooser if you're on a Snapdragon X / Surface Pro. The Intel build runs through Microsoft's x64-on-ARM emulator, which works but adds cold-start delay and battery drain. Gemini's
grpcio dependency still has no ARM64 wheel for Python 3.13, so Groq is the default working transcription path on ARM.Why isn't it just a tray icon?
It is. The window is optional — close it any time and Murmur keeps running silently. start.bat launches without a console; the tray icon is the persistent presence.
Is there a Mac or Linux build?
Not yet. Murmur is built around Windows-specific keyboard hooks. A Mac build is on the roadmap but not promised.
Will the accuracy review send my audio anywhere new?
No. The review never sees your audio. It sends only the transcribed text to the chat model of the provider you've already configured — same Groq or Gemini key, same endpoint as your transcription. No new vendor, no new traffic destination.
Can I turn the AI features off entirely?
Yes. Set Accuracy review to
never and Format mode to verbatim, and Murmur will type back exactly what the speech model returned — no second pass, no smart formatting, no rewrites.