⌖
Choose the destination
Click the email, document, browser form, editor or terminal field where the text belongs.
Windows dictation · v1.9.2
Speak anywhere.
Type everywhere.
Private, low-latency speech-to-text for Windows. Focus a field, hold a shortcut, speak, then release. Your finished words arrive where you were working.
Windows 10 & 11 · x64 desktop · Models downloaded separately · No account
ToProject team
Launch notes
Morning all,
The revised Windows build is ready for review. I’ve included the model notes and updated installation steps.
The revised Windows build is ready for review. I’ve included the model notes and updated installation steps.
●Listening
Ctrl + Shift · release to finish
Ctrl + Shift · release to finish
PS C:\work> git status
On branch release/1.9.2
PS C:\work> update the release notes_
On branch release/1.9.2
PS C:\work> update the release notes_
✓Local speech processing
✓Designed for Windows fields
✓Choose your local model
✓No account required
One shortcut. One thought.
01 / FOCUS
02 / SPEAK
03 / RELEASE
Dictation that stays out of your way.
Blabbermouth captures the field you chose before recording, does the transcription on your PC, then delivers the completed result.
◉ CtrlShift
Hold or toggle
Use your global shortcut. The compact overlay and pulsing tray icon confirm recording.
✓
Continue working
Release or toggle off. The finished local transcription is sent to your captured target.
Local is the architecture
Your voice stays on your machine.
Normal dictation does not send microphone audio or transcripts to a cloud transcription service. Text history is optional and local. Recovery audio stays in memory unless you explicitly save a WAV.
◉Your microphoneCaptured while your shortcut is active
→
▣Local modelWhisper or Parakeet runs on your PC
→
⌨Target fieldCompleted text delivered in Windows
THIS ENTIRE PATH RUNS ON YOUR WINDOWS PC
Built for repeated work
⌨
Aa
↳
☷
CC
From raw speech to useful text.
Dictate your way
Hold-to-talk or press once to start and again to stop. Choose a normal key combination or a silent modifier-only shortcut that won’t leak a printable character into terminals.
ACTIVE SHORTCUT
Ctrl + Shift
● Ready · model kept loaded
Ctrl + Shift
● Ready · model kept loaded
- Optional silence completion
- Microphone selection
- Escape to cancel
- Clear sounds and recording feedback
Make it yours
Custom vocabulary, deterministic corrections, snippets, filler removal and spoken punctuation.
App-aware profiles
Choose formal, casual or verbatim output by destination process.
Local text history
Optionally keep 50 transcripts, search favourites, edit, learn and retype. Audio is not added to history.
Files and subtitles
Batch common audio and video files, then export SRT, WebVTT or sidecar transcripts.
Optional model downloads
Whisper · MultilingualSmall~466 MiB
Whisper · MultilingualMedium~1.5 GiB
Whisper · MultilingualLarge v3~2.9 GiB
Parakeet · 25 European languagesTDT 0.6B v3 INT8~465 MiB
Install lean. Choose the engine that fits.
The MSI does not include model weights. Download a model after installation with its storage size visible before you commit.
The lighter Whisper choice for everyday local dictation.
A larger option for users who favour accuracy.
The largest included Whisper catalogue choice.
A fast local alternative powered by sherpa-onnx.
Requirements
- Windows 10 or Windows 11, x64 desktop
- Local storage for one or more optional models
- Whisper can use supported NVIDIA CUDA, with CPU fallback
- Parakeet currently runs on CPU
Good to know
- Text arrives after recording, not word by word while speaking
- Compatibility varies by destination control
- Elevated targets require Blabbermouth to run elevated too
- A model must be downloaded before recognition can work
Blabbermouth 1.9.2
Give your keyboard a quieter job.
Local Windows dictation, ready from the system tray whenever you need it.
[DOWNLOAD URL PLACEHOLDER] · [PRICING / LICENCE PLACEHOLDER] · [INSTALLER SIGNING DETAILS PLACEHOLDER]