100% local · no subscriptions
SweetWhisper turns voice into text 3–4× faster than you type: hold a key, speak — and the finished text appears in Telegram, Word, email or code. Everything is recognized on your PC, no internet needed.
Features
Speech is ~120 words per minute, typing is ~40. Dictate text into messengers, email, documents and code.
Nothing to copy: recognized text is automatically inserted into the active window — Telegram, Word, your browser or IDE.
Will the contract be ready by tomorrow?
Yes, I’ll send the final version by noon — just two clauses left to align with legal.
dictated in 4 seconds
The best recognition models inside, including GigaAM purpose-built for Russian speech. Everything downloads in one click, nothing to configure.
«please move the meeting to three p m»
Please move the meeting to 3:00 PM.
Whisper · 90+ languagesRemoves “umm” and filler words, fixes punctuation, formats and translates. Runs right on your PC — no internet required.
“uh in the supply agreement we need to change the payment term from thirty days to forty five and also add a clause about a zero point one percent daily penalty”
Supply agreement: change the payment term from 30 to 45 days; add a clause on a 0.1% daily penalty.
With a graphics card recognition takes a fraction of a second; without one it runs on the CPU. NVIDIA, AMD or Intel — any Windows 10/11 laptop works.
Say “Kitty, undo” — and the app undoes your last action. Fully customizable: key chords, launching programs.
Every transcription is stored locally: search, copy, see your stats. Nothing is sent to the cloud.
How it works
One configurable key: hold it — speak, release — done. The floating bar stays on top, always at hand.
Voice activity detection trims the silence. The model transcribes right on your GPU.
The result is instantly inserted into the active input field — in any Windows app.
Interface
Model, speed, hotkeys and text insertion — all in one place. The app checks your microphone, model and hotkey on its own and tells you when everything is ready for dictation.
Privacy
SweetWhisper is built local-first: audio is recorded, transcribed and discarded on your PC. No accounts, no telemetry, no “analytics to improve the service”. Cloud engines (OpenAI, Yandex) are strictly opt-in, and their keys live in the protected Windows credential store.
Comparison
| Comparison | SweetWhisper | Windows dictation (Win+H) | Cloud services |
|---|---|---|---|
| Works offline | ✓ | ✗ for most languages | ✗ |
| Audio stays on your computer | ✓ | ✗ | ✗ |
| Russian speech quality | GigaAM — built for Russian | average | good |
| AI text cleanup | ✓ Pro, on-device | ✗ | in the cloud |
| Payment | free · Pro by donation | free | monthly subscription |
If Win+H is enough and the cloud doesn’t bother you — use it. SweetWhisper is for people who need accuracy, privacy and offline work.
Pricing
SweetWhisper is freemium: full dictation is available right away, forever. Pro is a one-time donation that supports development — with AI bonuses as a thank-you.
Free
1 490 ₽
one-time donation · forever
As a thank-you for your support:
Free vs Pro in detail
| Free vs Pro in detail | Free | Pro |
|---|---|---|
| Unlimited local dictation | ✓ | ✓ |
| All recognition models and GPU speed | ✓ | ✓ |
| Voice commands, history, stats | ✓ | ✓ |
| AI cleanup, punctuation and formatting | — | ✓ |
| Task modes (writing, documents, code) | — | ✓ |
| Local AI — text never leaves the PC | — | ✓ |
| Cloud AI of your choice | — | ✓ |
| Meeting transcription | — | soon |
FAQ
Yes. Speech recognition runs locally on your computer — you only need internet once, to download a model. Cloud engines (OpenAI, Yandex SpeechKit) are optional; the default is fully offline.
Whisper recognizes 90+ languages including English and Russian. For Russian there is a dedicated GigaAM model with excellent accuracy. The app UI is available in English and Russian.
Whisper large-v3-turbo handles everyday speech, terms and mixed-language sentences very well, and for Russian the dedicated GigaAM model (trained on thousands of hours of Russian speech) is noticeably more accurate than universal models. Punctuation and capitalization are added automatically.
Windows 10/11, a graphics card is optional: with one recognition is faster (any NVIDIA, AMD or Intel works), without one everything runs on the CPU. Recognition only runs while you dictate — the rest of the time the app just waits for the hotkey. Models take 500 MB – 3 GB of disk space.
No — unless you turn that on yourself. By default Pro uses the built-in local AI: text is processed on your PC and never leaves it, which is safe even for confidential documents. Cloud models are connected only manually, with your own key.
Built-in dictation sends your voice to the Microsoft cloud and needs internet for most languages. SweetWhisper recognizes locally and offline, is more accurate for Russian (GigaAM), inserts text into any app and can polish the result with AI (Pro).
SweetWhisper is freemium: all core dictation is free, unlimited, forever. Pro is a one-time donation that supports development — AI text processing and other bonuses unlock as a thank-you. No subscriptions.
Nowhere. Audio is processed on your PC and never leaves it. Transcription history lives in a local database you can clear at any time.
Speed (no network latency), privacy (audio never leaves your machine), cost (no per-minute fees) and it works anywhere — even on a plane.
Linux support is in development: core functionality already builds and runs, a full release comes later.
Installer or portable build — no sign-up, no accounts.
Release in the works. Leave a contact — we’ll send you the link first.
Windows 10/11 · graphics card optional
