Place your cursor
Open a regular text field in a message, document or note.
FREE TO USE. NO SUBSCRIPTIONS.
Your voice —
your text.
Dictate messages, documents and notes into a regular text field. VibeScribe recognizes speech on your computer and inserts the text when you release your hotkey.
Windows 10 / 11 · x64
HOW IT WORKS
Hold → release → text
Scripted example. No microphone access.
Choose an example type and hold the button. Release it to see the result.
The example text will appear here…
01 / THREE STEPS
How to dictate with VibeScribe
Hotkey
Hold your hotkey
Wait for “Listening” and say what’s on your mind. You can change the shortcut.
VibeScribe
Release your hotkey
Keep the same field focused until recognition finishes. Review the inserted text and send it yourself.
02 / FILLER WORDS AND REPETITIONS
Notice your speech habits
We do not remove recognized filler words or repetitions.
Removing them automatically is technically possible. We deliberately keep them so you can notice your speech habits and practise speaking more clearly. This applies to every model.
Read the transcript, notice an unnecessary repetition and try saying the sentence again. You decide what to practise; the app does not rewrite what you said.
Whisper can make mistakes and omit words. Verbatim transcription is not guaranteed. VibeScribe does not score pronunciation — you reflect on your own speech.
VibeScribe
“I need to, need to write this down.”
An example of a repetition you may notice when reading.
The recognized repetition stays in the text.
03 / FEATURES
Speech recognition and text insertion
Recognition on your PC
Recognition on your computer. No audio or transcripts sent to the cloud, no account. Once a model is downloaded, dictation works offline.
You control insertion
Switched windows during recognition? Your text stays in VibeScribe for copying. The app never presses Enter for you.
CPU or NVIDIA
Save memory or keep the model ready for a faster response. Use your CPU or a compatible NVIDIA graphics card.
APPEARANCE
Adjust the widget
Choose its colour, size and transparency. The widget shows dictation status without taking focus away from your work.
04 / CHOOSING A MODEL
Models for English and Russian
Choose from three models in the app. Turbo Q5 and Small Q5 support both languages; Base English supports English only. Speed depends on your computer.
RECOGNITION MODELRU + EN
Turbo Q5
A model from the Whisper large-v3-turbo family for English and Russian speech. Speed depends on your processor and graphics card.
- Languages
- English / Russian
- Download
- 574 MB
- Device
- CPU / NVIDIA
RECOGNITION MODELRU + EN
Small Q5
For English and Russian speech. The 190 MB download is smaller than Turbo. Compare recognition quality and speed using your own phrases.
- Languages
- English / Russian
- Download
- 190 MB
- Device
- CPU / NVIDIA
RECOGNITION MODELEN
Base English
A compact model for English dictation. For Russian speech, choose Turbo Q5 or Small Q5.
- Languages
- English
- Download
- 148 MB
- Device
- CPU / NVIDIA
NVIDIA CUDA is an optional download of about 675 MB. CPU recognition does not need it. A compatible driver and enough GPU memory are required.
Inside the app
Open the full-size image
FREE. NO SUBSCRIPTIONS.
Download VibeScribe
for Windows
There are no payments for the app or its use. Choose the installer for a regular installation, Portable to run from a folder, or the offline bundle for a PC without internet access.
Windows installer
≈ 152 MBFor a regular installation on your PC. Download a model from the app on first launch.
Download installerPortable ZIP
≈ 83 MBExtract the archive to a separate folder and run VibeScribe.exe. Models are downloaded separately.
Download PortableOffline bundle
≈ 1.61 GBFor a PC without internet access: the app, three models and CUDA. Component preparation needs at least 5 GB in the data folder.
Download offline bundleWindows 10/11 x64 · 8 GB RAM or more recommended · Microphone
All versions and release notesApproximate download sizes for version v0.6.4 · 2026-09-10.
The app is not yet publisher-signed: Windows may display a warning. Checksums are in SHA256SUMS.txt on each release page. A new PC may require Microsoft Visual C++ Runtime x64.
QUESTIONS AND ANSWERS
Before you install
How much does VibeScribe cost?
VibeScribe is completely free to use. No subscriptions or payments — you do not pay for the app or for using it. Just download it, choose a model and get started.
Do I need an internet connection?
Not for dictation, once a model is installed. Internet is used to download the app, models and CUDA, and check for updates. Automatic checks can be disabled; downloading and restarting require your action.
Where do my recordings go?
VibeScribe recognizes speech locally. Temporary audio files are deleted after processing; a crash can leave a file behind. History is off by default. If enabled, transcripts are stored on your computer. Clipboard insertion is subject to Windows clipboard history and sync settings; Unicode input without the clipboard is also available.
Why keep filler words?
It is a deliberate principle for every model: the app applies no filler-word filter or editorial rewriting. Reviewing a transcript can help you notice speech habits. Recognition itself can still omit words — exact verbatim transcription is not guaranteed.
Do I need an NVIDIA graphics card?
No, the app works on your CPU. Compatible NVIDIA hardware with the optional CUDA component can accelerate recognition. AMD and Intel graphics use CPU recognition. Speed depends on the model and hardware.
Does it work in every application?
VibeScribe is designed for regular text fields. Password fields, protected fields and applications with higher privileges are unsupported. Some editors and remote sessions may reject automatic input. In that case, copy the result manually.
Can I use Russian too?
Yes. English and Russian dictation are selected independently of the interface language. Turbo Q5 and Small Q5 support both; Base English is for English only.