Vox0 · free offline dictation for Windows
Offline
dictation.
Hold a key and speak. Your words appear in the app you're using. A free, open source alternative to Wispr Flow, with transcription running 100% locally on your PC, built for NVIDIA graphics cards with CPU fallback. Download the speech model once, then dictate offline.
No actors, no script -- just Bryan talking to his own PC, an RTX 3070, which is why it's this fast.
NVIDIA support included, with CPU fallback. A compatible NVIDIA graphics card and a current driver are recommended.
Current release: Vox0 0.3.6 · Windows 10 and 11 · nothing else to install
Step by step install guide· All files and release notes· SHA256 checksums
Automatic language detection is back: switch languages between dictations.
Mixing languages in one recording can cause errors. A fixed language can improve accuracy, especially on short clips.
How it works
Three things, and none of them is learning commands
- HOLDPress the keyHold Insert. The microphone opens only now, and closes the moment you let go.
- SPEAKSay it out loudTalk normally. Nothing to memorise, no wake word.
- RELEASELet goA second later the text is typed where your cursor was.
The speech recognition runs on your own computer usingWhisper. Your voice is never uploaded. You can pull out the network cable and it still works.
What it looks like
A small window that stays out of the way
How it compares
Vox0 and Wispr Flow: dictation, with a different approach
Both turn speech into text where you type. Vox0 is for Windows users who want their voice processed entirely on their own PC.
| Feature | Vox0 | Wispr Flow |
|---|---|---|
| Price | Free, no subscription | Free and paid plans |
| Transcription | On your PC, using Whisper | In the cloud |
| Works offline | Yes, after downloading the model | No, internet required |
| Dictation into apps | Yes | Yes |
| Platforms | Windows 10 and 11 | Mac, Windows, iOS and Android |
| Text processing | Transcription, punctuation and number formatting | AI cleanup and rewriting |
Vox0 focuses on local dictation. Flow also offers AI rewriting and cross-device features. Compare those features with your needs.
Sources: Wispr Flow features,cloud transcription and internet requirement,plans. Checked .
Questions
The ones people actually ask
Does it work offline?
Yes. The first run downloads the speech model, about 460 MB. After that it never needs the internet again, and your voice is processed entirely on your own machine.
Is it free?
Yes, and it is open source under the MIT license. No account, no subscription, no paid tier, no trial that runs out.
Is Vox0 a free offline alternative to Wispr Flow?
Yes, for dictating text into Windows apps. Vox0 runs Whisper entirely on your PC and works offline after a one-time speech model download. Wispr Flow transcribes in the cloud and needs an internet connection. Vox0 focuses on dictation, punctuation and number formatting; it does not offer Flow's AI rewriting or cross-device features.
How is it different from Windows Voice Typing (Win+H)?
Windows Voice Typing needs an internet connection. Vox0 runs Whisper on your own PC and works offline after downloading the speech model. You can choose the model according to your hardware and accuracy needs.
Which languages does it support?
Spanish, English, Portuguese, French, German and Italian. Automatic mode detects the language for each dictation. You can also choose a fixed language to improve accuracy on short clips. Mixing languages in one recording can cause errors.
Do I need a graphics card?
Vox0 is designed for a compatible NVIDIA graphics card with a current driver. The installer includes its NVIDIA runtime and CPU fallback. You can also select CPU manually in Settings.
Is my voice sent anywhere?
No. Nothing leaves your computer. The microphone is only open while you hold the key, so nothing is listening the rest of the time.
Which key does it use?
Insert by default, and you can change it to any key or an extra mouse button. Avoid normal letters: the key is not blocked, so choosing q would start a recording every time you typed one.