If typing is slower than thinking
AuroraWhisp turns voice into text in under a second. Think, speak, release the key — and keep going. No "wait, I need to type that out."
Voice typing that doesn't eavesdrop.
Press a key, speak, release — text appears in any window. Recognition runs on your computer. No cloud. No account.
Sound familiar?
If any of this sounds like you — you're in the right place
Five things we keep hearing from early users.
If your wrists ache by the end of the day
Bind any hotkey you like. No menus, no per-session setup. A keyboard alternative that doesn't hurt.
If Windows built-in dictation drives you crazy
Modern recognition models tuned for English — and nine other languages. Punctuation handled. Names and jargon you can teach.
If you don't want the cloud listening
Recognition runs on your computer. Nothing leaves your machine. No accounts, no telemetry — nothing to leak.
If Wispr Flow is too expensive
There is a free tier — most people are fine with it. Pro is for those who need more: unlimited words and the heavy Whisper Medium / Large models.
How it works
Three steps. No training, no setup.
From "opened the app" to "dictated a paragraph into Word" — under a minute after install.
1
Press
Any hotkey you like — Ctrl+Space, F9, right Ctrl, Caps Lock. Set it once, you're done.
2
Speak
The widget shows what it hears. Speak naturally — pauses, inflections, mid-sentence corrections — the model handles it.
3
Release
A second later, the text appears where your cursor is. Word, Slack, your IDE, the browser — anywhere.
One hotkey, any window
Ctrl+Space, F9 or any key you bind — dictate into any field, in any application. No per-session setup.
Your voice stays with you
Recognition runs on your machine. Not sent anywhere — not to a server, not to "anonymous logs." Nothing to leak.
Fifteen languages out of the box
English, Spanish, German, French, Italian, Portuguese, Dutch, Polish, Czech, Turkish, Russian, Ukrainian, Japanese, Korean, Chinese. Plus an "Auto" mode — the app detects the language from your audio.
A widget that doesn't get in the way
Small, appears only when you speak. Twelve appearance options — pick the one that feels right.
Learns your words
Coworker names, project jargon, unusual words — record once by voice, the app remembers every variant the way it hears you.
150 ms from thought to text
A 5-second phrase recognised in ~150 ms on a regular CPU — faster than you can lift your finger off the key. Heavy Whisper Medium/Large models on a GPU run in 100–300 ms.
Instantly in your language
On first launch the app reads your Windows system locale and loads the right model. German user downloads the .exe → app starts in German, no setup. Same for the UI itself — it ships in your language (RU/EN/DE/ES/FR/IT).
Picks the right model for your hardware
The app detects your CPU, RAM and GPU and suggests an optimal model. Weak laptop → Whisper Base. Strong machine with CUDA → Whisper Large-v3 in 100–200 ms. No "try this, then try that" trial and error.
Audio and video transcription
Drop an mp3 / mp4 / m4a / wav onto transcribe.bat — get a ready txt next to it and a copy in your clipboard. Transcribing a one-hour lecture takes a minute or two.
Voice autoreplace
Say "comma" — get a comma. 19 built-in rules for punctuation and formatting plus your own. Numbers ("two thousand five hundred" → 2500) and auto-capitalisation work out of the box.
How it looks
The app remembers everything you dictate
The stats dashboard shows how many words you dictated today, in total, and at what speed. Everything is stored locally in SQLite on your computer.
How it works
Three seconds from thought to text
Press and hold a key — while held, the app listens. Release — and a moment later your speech appears wherever your cursor is. Any field, any application.
Hold Ctrl+Space → speak → release → done
live animation · reacts to voice
Fifteen languages
Speak in your own.
Recognition works at the same high quality across fifteen languages. One click in the tray to switch — no restart.
English Español Deutsch Français Italiano Português Nederlands Polski Čeština Türkçe Русский Українська 日本語 한국어 中文
Comparison
How we differ from Wispr Flow and Win+H
Short version: local, multilingual, no subscription. Full tables with five competitors live on the dedicated comparison pages.
| App | Price | Where it runs | No subscription |
|---|---|---|---|
| AuroraWhisp | Free + Pro $19.90 | Your machine | Yes |
| Wispr Flow | $15/mo | Cloud | No |
| WhispeRu | Paid | Your machine | Yes |
| Win+H | Free | Your machine* | Yes |
| Dragon | $300+ | Your machine | Yes |
Privacy
Voice is deeply personal.
When you speak, you say things you'd never type publicly. Private messages, work passwords, thinking out loud, salary numbers, medical questions. AuroraWhisp doesn't send any of that to the cloud, doesn't store it on servers, and doesn't use it to train models.
Recognition on your computer
Not a single byte of audio leaves your machine. Verify it in Wireshark, or block the app at your firewall — it will keep working.
No accounts
Download, use it. No email at install, no phone confirmation, no signup at all.
No telemetry
Optional anonymous stats — only if you turn them on yourself in settings. Off by default. No audio, no text, no device IDs.
Plain-language policy — no traps
Privacy policy written in human language. If anything ever changes, we'll tell you in advance. The version of the app you installed will keep working under the rules in effect when you installed it.
Pricing
Free now. Pro later.
Free — 10,000 recognised words per day, all 15 languages, light models. Pro — $19.90 once: unlimited words, heavy models (Whisper Medium / Large-v3), up to 3 devices, no-questions 14-day refund. No subscription on principle — buy once, yours forever.
Free
$0
Forever. Free — 10,000 words/day; Pro — unlimited.
Pro
$19.90
Founders price. One-time, lifetime. All future updates included.
FAQ
Most common questions
Does anything really not go to the cloud?
Really. Recognition runs through local models on your computer. You can verify in Wireshark, or block the app at your firewall — it will keep working.
What are the system requirements?
Windows 10 (build 1809+) or 11. 4 GB RAM for the base model. ~500 MB disk. NVIDIA GPU optional — gives 5–15× speedup for medium/large models.
How does it differ from Wispr Flow?
Wispr is cloud-based, $15 a month, AI auto-edits your speech. We are local, with a free tier forever, edits are minimal. We also have a proper Russian model.
When Mac or Linux?
Planned, no firm date. Subscribe in the Contact section — you'll hear first.
What if you shut down in a year?
The app keeps working offline as long as Windows runs. Pro features won't disable, since the license is checked locally. Whether we shut down or not — your installed version is unaffected.