01 / FOCUS
Start where the words belong.
Click an editable field in email, a browser, a document, chat, an editor or a terminal. Blabbermouth remembers the Windows target.
Subject: Thursday's review
Write a message… │
Local dictation for Windows 10 & 11
Speak anywhere.Type everywhere.
Hold a shortcut, speak, release. Blabbermouth turns your voice into finished text in the Windows field you were using, with transcription kept on your PC.
Version 1.9.2 · Windows 10/11 x64 · Speech model downloaded separately
Ctrl + Shift
Maya Thompson
Hi Maya,
The revised build is ready for you to try this afternoon.│
The revised build is ready for you to try this afternoon.│
Local processingWorks across Windows appsChoose your modelNo account required
One continuous action
Focus. Speak. Release.
Blabbermouth captures the destination first, listens while your shortcut is active, then delivers the completed result. No dedicated editor. No transcript stream pretending to be final.
02 / SPEAK
Hold, toggle, or wait for quiet.
Use your chosen shortcut. A compact topmost overlay and pulsing red tray icon make recording obvious. Escape cancels immediately.
●
Ctrl+Shiftmodifier-only
03 / RELEASE
Receive finished text.
Blabbermouth transcribes on your machine, applies your enabled text rules, and sends the result back to the field you focused.
The revised build is ready for you to try this afternoon.
✓ DELIVERED LOCALLY
Local by design
Your voice takes the short route.
Normal dictation does not send microphone audio or transcripts to a cloud transcription service. No account is required.
Settings, optional text history, models and statistics stay under your control in local application data.
MicrophoneCaptured while your shortcut is active
→
Local modelWhisper or Parakeet on this PC
→
Focused fieldFinished text delivered in Windows
This entire path stays on your PC
Make it yours
Less fixing. More flowing.
Tune how Blabbermouth starts, what it recognises, and how your finished words arrive.
A shortcut that stays out of the way
Choose hold-to-talk, press-once toggle, or optional silence completion. Modifier-only shortcuts can stay silent in terminals.
- Hold or toggle
- Custom global shortcut
- Mic choice
- Audio cues
Teach it the words you use.
Condition Whisper with specialist terms and proper nouns, then repair recurring errors with deterministic replacements.
- Custom vocabulary
- Corrections
Speak once. Expand.
Exact triggers insert reusable text, dates, times, new lines and tabs inside a sentence.
Finished means finished.
Optional punctuation, filler removal, trailing spaces and voice commands prepare the result before delivery.
Match the destination.
Choose formal, casual or verbatim output by destination process, and optionally say “send it” to submit.
Optional downloads
Choose your local listener.
The installer stays lean. Pick a model after installation, see its storage size first, and change later.
Small
The lighter multilingual Whisper option for everyday local dictation.
~466 MiBMultilingualMedium
A larger Whisper option for users who favour accuracy.
~1.5 GiBMultilingualLarge v3
The largest included Whisper catalogue choice.
~2.9 GiBMultilingualTDT 0.6B v3 INT8
A fast local alternative powered by sherpa-onnx.
~465 MiB25 European languagesⓘWhisper can optionally translate supported speech into English and can use supported NVIDIA CUDA. Parakeet currently runs on CPU and transcribes without translation.
After the sentence
History when useful. Audio only when asked.
Keep up to 50 text transcripts locally, or leave history off. Recover the last recording from memory, then save a WAV only if you choose.
HistorySearch transcripts…
★The revised build is ready for you…10:42
☆Can we move the review to Thursday?09:18
☆npm run build and then open…Yesterday
TEXT ONLY · MARKDOWN / JSON EXPORT
Files2 in queue
▤interview-01.mp3COMPLETE
▤research-call.m4aPROCESSING
＋Drop WAV, MP3, M4A, AAC, WMA, MP4 or MOV
OPTIONAL SRT · WEBVTT · SIDECAR TEXT
Fits your Windows day
Present in the tray. Ready at the cursor.
Start with Windows, follow your system theme, keep your selected model prepared, and export your preferences as JSON.
01Windows 10 or Windows 11, x64 desktop
02Per-user installation and optional launch at sign-in
03Light, dark and High Contrast system themes
Blabbermouth 1.9.2
Put your voice where the cursor is.
Private, local speech-to-text for Windows 10 and 11. Download the lean installer, then choose the model that suits you.
Pricing & licence: [PRICING / LICENCE PLACEHOLDER]