Free · Open source · 100% on-device
Dictate anywhere. Nothing leaves your machine.
AuraScribe turns speech into clean, punctuated text in any Windows app. Tap a hotkey, talk, and it lands the moment you stop. Every word is transcribed on your own PC, so there is no account, no subscription, and nothing sent to a server.
~8.6 MB · Windows 10 and 11 · MIT licensed
Why it exists
What if it never left your machine?
Languages
And the scripts the fast tools skip.
Ship the release notes this afternoon.
Real-time on a plain CPU.
65+
languages, every one of them on your device.
- 25
- European languages
- NVIDIA Parakeet
- ~40
- Asian languages
- Dolphin
- 2
- Indic: Malayalam, Kannada
- AI4Bharat IndicConformer
- 1
- English, tuned for speed
- Moonshine
Each engine detects the language itself inside its own region, so there is no picker to set and no round trip to a server. Download only the ones you actually speak.
Free forever
It runs on your machine. Full stop.
Download the installer, set a hotkey, and start dictating in under a minute. No account to create, nothing to pay, and no cloud to trust with your voice.
Built on open work
AuraScribe does not train its own models. It credits the people who did.
- Moonshine
- English speech recognitionby Useful Sensors
- Parakeet TDT
- 25 European languagesby NVIDIA
- Dolphin
- About 40 Asian languagesby DataoceanAI and Tsinghua
- IndicConformer
- Malayalam and Kannadaby AI4Bharat
- sherpa-onnx
- On-device inference runtimeby k2-fsa
- ONNX Runtime
- Model executionby Microsoft
- Tauri
- The 8.6 MB application shellby Tauri contributors
How it works
Three keys, then just talk.
There is no window to open, no record button to find, and no upload waiting to finish. The whole loop happens where your hands already are.
Press
Hit Ctrl + Shift + Space in whatever window you are already in. AuraScribe is listening straight away, even with its own window closed, so nothing interrupts what you were doing.
Rebindable to any combination
Speak
Talk the way you would to a person. Transcription runs on your own CPU while you speak, and the cleanup pass fixes punctuation and casing and drops the ums and uhs as it goes.
Runs faster than real time
Release
The finished text is typed into the app that had your cursor. No copy, no paste, no switching windows, and no dialog asking you to confirm anything.
Works in any Windows app
What you get
Small, quiet, and completely yours.
Your voice never leaves the machine.
Audio is transcribed on your own CPU, and the cleanup is ordinary local text processing rather than a model in someone else’s data centre. The single network request the app ever makes is downloading a speech model, once. No account, no telemetry, no analytics.
0.1×
Real time on a plain CPU. A ten second sentence is transcribed in about one, so the text is already there when you stop.
Anywhere you can type
Browsers, editors, terminals, chat boxes, form fields. It types where your cursor already is.
Tidied as you speak
Punctuation, capitals and filler words are handled locally, on every engine. Toggle it off if you would rather have the raw transcript.
Your words, your spellings
A personal dictionary fixes names and jargon, and snippets expand a spoken phrase into whatever boilerplate you keep retyping.
Every transcript stays on your disk.
Searchable history in a local SQLite file you can delete by date range whenever you like. Nothing is synced, because there is nowhere for it to sync to.
The honest comparison
A free, open alternative to Wispr Flow, Superwhisper, and Dragon.
Most dictation tools bill monthly, need an account, or stream your voice to a server you do not control. This is where AuraScribe differs, and where it does not. Only claims we can stand behind.
| AuraScribe | Wispr Flow | |
|---|---|---|
| Price | Free forever | Subscription |
| Open source | ||
| Runs offline (no cloud) | ||
| No account required | ||
| Types into any app | ||
| Malayalam / Kannada | Cloud only |
Each tool in its default configuration. Windows Voice Typing sends audio to Microsoft unless you disable online speech recognition.
Questions
The things people ask first.
- Is AuraScribe really free?
- Yes. AuraScribe is free forever and open source under the MIT license. There is no paid tier, no subscription, and no account to create.
- Does my voice get sent to the cloud?
- No. Every word is transcribed on your own PC. The only network request the app ever makes is downloading a speech model once, after which it works fully offline.
- How is it different from Wispr Flow, Superwhisper, or Dragon?
- AuraScribe is free, open source, and 100% offline, with no account required. Wispr Flow and Superwhisper are paid and cloud-based; Dragon is expensive. It is the private, local alternative.
- Which languages does it support?
- English plus 25 European languages, roughly 40 Asian languages, and Malayalam and Kannada, using four on-device engines. Each auto-detects the language within its region.
- What are the system requirements?
- Windows 10 or 11 (64-bit) and a microphone. The installer is about 8.6 MB and runs faster than real time on an ordinary CPU. No internet is needed after the first model download.
- Does it work in every application?
- Yes. Press the hotkey in any window and the finished text is typed straight into whatever app has your cursor: browsers, editors, terminals, chat boxes, and form fields.
Get in touch
Questions, bugs, or ideas.
Found a bug or want a language added? Open an issue on GitHub, or send a note here and it reaches the maintainer directly.
contact.aurascribe@gmail.com