Voice-to-text for Windows that runs fully on your machine — or with your own API key. Hold a hotkey, talk naturally, watch perfect text appear in any app.
Pastes anywhere you can type — no integrations needed.
A glassy capsule glows above your work. The moment you let go, polished text drops into the focused app.
The fastest path from thought to text on your PC.
Press and hold your hotkey — Ctrl + Win by default. Or pick any other combo.
Talk naturally while the aurora pill glows. Filler words, pauses, restarts — just speak.
Polished text is transcribed and pasted instantly into whatever app you were using.
Run Whisper entirely on your PC for free, or plug in your own API key for state-of-the-art accuracy. We never see your audio.
Audio never leaves your computer. Models cache locally on first download. Zero accounts, zero subscriptions.
Use any provider you already pay for. Keys are stored locally and never transmitted to WisprLocal servers — because there are none.
gpt-4o-transcribe
whisper-large-v3-turbo
nova-3
universal-2
Switch engines per-mode. Use local for private notes, cloud for high-accuracy work — same hotkey.
A natural speaking pace clocks in around 4–5× faster than the average typist. WisprLocal turns that gap into hours back in your week.
Typing
45wpm
average keyboard speed
WISPRLOCAL
220wpm
your natural speaking pace
≈ 4.9× faster
Audio is transcribed on-device. No internet required, no servers in the loop, nothing leaves your machine.
Optional AI pass removes filler words, fixes self-corrections, and adds punctuation — your transcript reads like writing.
Add names, jargon, and acronyms organized by category. The words that matter to you come out right every time.
Ctrl + Win, Ctrl + Fn or any combo. Hold to speak, release to insert. Never blocks the keys for normal use.
A glassy capsule glows while you dictate, shifting state through recording, transcribing, and inserting — then quietly shrink.
Plug in OpenAI, Groq, Deepgram, or AssemblyAI for cloud-grade accuracy. Keys stay on your machine, billing stays with you.
Optionally launches at startup and stays out of the way until you press your hotkey. Wakes instantly, every time.
Drops your transcript into the focused app — emails, code editors, chat, browsers, terminals, forms. No integrations.
Multi languages supported out of the box. Switch on the fly or let the model auto-detect — no setting changes.
An optional AI pass cleans your dictation as it lands — fillers, stutters and self-corrections vanish, punctuation appears, and what gets typed reads like you wrote it.
An optional AI pass scrubs filler words, repairs self-corrections, and lands clean punctuation — so what gets typed reads like writing.
What you said
um so I was thinking maybe, uh, we should send the email today, no actually first thing tomorrow morning
What got typed
I was thinking we should send the email first thing tomorrow morning.
01 · Fillers
The little words you say but never want to write.
02 · Punctuation
Speech becomes a sentence that breathes — without you ever saying "comma" or "question mark".
what time is the meeting tomorrow
What time is the meeting tomorrow?
03 · Tone
Switch modes for casual messages, formal email, or code-comment style. Same voice, different polish.
03 · Self-corrections
Smart Typing keeps the version you actually meant.
What you said
Send the deck today, no actually, tomorrow morning before the standup.
Send the deck tomorrow morning.
Telemetry never includes the words you said or the words we cleaned.
Polish on for high-stakes writing. Off when you want the raw transcript.
Teach WisprLocal the names, jargon, and acronyms you use every day — organized by category so it knows the right meaning, every time.
A glassy capsule sits above your work and shifts color to show what's happening — recording, transcribing, polishing, inserting. No popups, no focus stealing.
A clear, honest comparison vs. the alternatives — so you can decide for yourself.
| Capability | WISPRLOCAL | Cloud dictation services | OS built-in dictation |
|---|---|---|---|
| 100% on-device option | ● | ○ | ◐ |
| Bring your own API key | ● | ○ | ○ |
| Free forever | ● | ○ | ● |
| No account required | ● | ○ | ● |
| Sub-second latency | ● | ◐ | ◐ |
| Choose your own model | ● | ◐ | ○ |
| Custom hotkey, push-to-talk | ● | ◐ | ◐ |
| Pastes into any focused app | ● | ◐ | ● |
| AI auto-edits / smart typing | ● | ◐ | ○ |
| Open source | ● | ○ | ○ |
Yes. The app is free and open. The local engine runs Whisper models on your machine at zero cost. If you opt into cloud providers, you pay them directly with your own API key — never us.
By default, no. Local mode runs entirely on your PC — audio never leaves your machine. If you choose to use a cloud provider (BYO key), audio goes directly from your computer to that provider; we never see it.
Windows 10 or 11, 64-bit. ~200 MB free disk space for the app, plus the model size you choose (75 MB to 3 GB). A microphone, obviously.
Start with Whisper Base — it's fast, accurate enough for most writing, and only ~150 MB. Move up to Small or Medium if you have a beefier machine. Use Large v3 for the best accuracy on long-form dictation.
Absolutely. Ctrl + Win is the default but you can pick any key or combo — Right Ctrl, Right Alt, F-keys, side mouse buttons, even a foot pedal. Hold-to-talk and toggle modes both work.
Multi languages out of the box, including English, Spanish, French, German, Mandarin, Hindi, Arabic, Portuguese, Japanese, Korean. Switch on the fly or let it auto-detect.
Every voice tool makes trade-offs. Some are cloud-only with great accuracy but a subscription and an account. Some are OS built-ins that are free but limited and clunky. WisprLocal's bet: be local-first and free, give you the cloud as an option via your own key, and stay out of your way.
Effortless voice dictation in every Windows application.
4× faster than typing, AI commands and auto-edits.