Microphone
Captured only while your shortcut or toggle is active.
Speak anywhere.
Type everywhere.
Private speech-to-text for the Windows field you are already using. Hold a shortcut, speak, release, and receive the finished text.
→
Local speech model
Whisper or Parakeet transcribes without a cloud transcription service.
→
Destination field
The completed text goes back to your captured Windows control.
THE WHOLE DICTATION PATH STAYS ON YOUR PC.Models are optional downloads after installation.
The everyday method
Three deliberate actions. Then carry on.
Blabbermouth stays in the system tray with your selected model prepared. Recording feedback is visible; delivery happens when the recording ends.
Reply to the team… |
Focus a field
Choose the email, document, form, editor or terminal control that should receive the text.
CtrlShift ●
Hold or toggle, then speak
A compact topmost overlay and pulsing red tray icon show when Blabbermouth is listening.
speech → “text”
Release and continue
Blabbermouth transcribes locally, applies your enabled rules, and sends the finished result.
Privacy, evidenced
Local is a behaviour, not a badge.
No sign-in is needed. Normal microphone audio and transcripts are not sent to a cloud transcription service.
Recognition on your PC
Your selected Whisper or Parakeet model runs locally. Built-in model downloads come from approved catalogues over HTTPS.
Text only, if you want it
Optional history keeps up to 50 text transcripts. Microphone audio is not added to that history.
Audio is not silently retained
The latest completed recording remains in memory for recovery and is written only when you choose Save WAV.
LOCAL DATA MAP%LOCALAPPDATA%\Blabbermouth
SettingsPreferences and app profilesLOCAL
HistoryUp to 50 text transcriptsOPTIONAL
ModelsWeights you choose to downloadLOCAL
Recovery WAVMemory-only unless savedEXPLICIT SAVE
AccountNo account or sign-in requiredNONE
The product
Small in the tray. Considered in use.
Configure once, then work across ordinary editable Windows fields with the activation style and text rules that suit you.
Control the moment before and after speech.
Blabbermouth handles activation, transcription and deterministic clean-up without forcing you into a dedicated writing app.
↗
Activation that fitsHold-to-talk, toggle, silence completion or silent modifier-only shortcuts.
⌘
Text productivityVocabulary, corrections, snippets, punctuation, filler removal and voice commands.
◎
Application-awareFormal, casual or verbatim profiles selected by destination process.
↺
RecoverableRetry the latest recording, search local history, edit and retype prior text.
Blabbermouth— □ ×GENERALPRODUCTIVITYMODELS & DATAHISTORY & FILES
Global shortcutCtrl + Shift · Hold to talk
Selected modelWhisper Medium · Ready●
Spoken punctuationConvert “comma”, “new paragraph”…
Text historyKeep the 50 most recent transcripts
Model choices
See the size before you download.
The installer stays lean. Choose and download local model weights after setup. No blanket speed or accuracy claims: engine and hardware trade-offs vary.
| Engine / model | Download | Language scope | Plain description |
|---|---|---|---|
| WHISPER Small | ~466 MiB | Multilingual | The lighter Whisper option for everyday dictation. |
| WHISPER Medium | ~1.5 GiB | Multilingual | A larger option for users who favour accuracy. |
| WHISPER Large v3 | ~2.9 GiB | Multilingual | The largest included Whisper catalogue choice. |
| PARAKEET TDT 0.6B v3 INT8 | ~465 MiB | 25 European languages | A fast local alternative powered by sherpa-onnx. |
Engine distinctions
Custom vocabulary conditions Whisper; do not assume the same mechanism applies to Parakeet. Translation into English is a Whisper capability. Parakeet currently transcribes on CPU and does not translate.
Beyond dictation
Useful after you stop speaking, too.
Searchable local history
Optionally keep, find, favourite, edit and export text transcripts as Markdown or JSON.
Batch file transcription
Drop WAV, MP3, M4A, AAC, WMA, MP4 or MOV files into a sequential local queue.
Subtitles and sidecars
Save timestamped Whisper segments as SRT or WebVTT, or a non-destructive transcript beside the source.
Requirements & boundaries
Know what you are installing.
Clear constraints make for a better fit. Blabbermouth is a Windows desktop tool designed for editable fields, not a promise of compatibility with every control.
System requirements
- Windows 10 or Windows 11, x64 desktop
- Storage for at least one optional local model
- Supported NVIDIA CUDA preferred for Whisper where available; CPU fallback
- Parakeet currently runs on CPU
Operating boundaries
- Finished text is delivered after recording, not word by word while speaking
- Compatibility varies with the destination control
- Elevated targets require Blabbermouth to run elevated too
- Recognition requires a downloaded model
Blabbermouth 1.9.2 / Windows x64
Download for Windows ↘Your words. Your PC. Your field.
[DOWNLOAD URL PLACEHOLDER] · [PRICING / LICENCE PLACEHOLDER] · [INSTALLER SIGNING & REPUTATION DETAILS PLACEHOLDER]