Press the hotkey
Press Ctrl+Shift+A in any app on your computer, and Voisee starts recording. You can change the shortcut in settings; Esc cancels the recording. On your phone, use the record button in the app.
Windows · macOS · Android
Control it with your voice
Voisee is hotkey dictation in Russian and other languages. Press, speak, and the text is already in the window you need, with periods, commas and capital letters.
0.1 s
minute of Russian speech*
≈3 s
minute of other languages*
60 min
Every week at no cost
* How long the server takes to recognise one minute of recording, not counting upload.
Dictation · Ctrl+Shift+A
How Voisee works
From keypress to text in your window in three steps.
The model recognises your speech
The recording goes to our GPU server. The model adds punctuation and capital letters, and if text cleanup is turned on in settings, it removes filler words and slips of the tongue.
The text lands where you type
Voisee pastes the text into the window where you started recording: an AI chat, an editor, a messenger. The text also stays on your clipboard and in the app history. The second hotkey pastes and sends right away.
What Voisee does
Dictation on desktop and mobile and speech recognition on our own GPU server: Russian is fast, other languages run in multilingual mode.
Russian and other languages
Fast mode recognises Russian speech: numbers come out as digits, and common English words like YouTube and HD in Latin script. For English and other languages there is multilingual mode, which is slower. In Auto mode the language is detected automatically.
Punctuation and capitals
The model adds commas, periods, question marks and capital letters itself, so the text arrives formatted rather than as one long line.
Text cleanup
Fixes spelling, grammar and recognition errors and writes terms the usual way: 1C, Excel, Docker. Removes filler words and slips of the tongue. The style adapts to the app you dictate into: messenger, email, code editor. It never writes for you and never switches how you address people from informal to formal. Works in the desktop and mobile apps. Turn it on in the app settings: by default the text arrives as recognised.
Command Mode: edit selected text by voice
Select text in any app, press Ctrl+Alt+E and say what to do with it: “shorten it”, “translate to English”, “make it more formal”. The result replaces the selection. Desktop app only.
On all your devices
On your computer, use the Ctrl+Shift+A hotkey: the Windows and macOS app lives in the system tray. On your phone, there is the Android app. For notes, there is an Obsidian plugin.
About voice typing →Obsidian plugin
Dictate at the cursor in a note, and keep your Voisee notes in two-way sync with your Obsidian vault.
History and sync
Everything you dictate is saved in the app history.
Note templates
Turn what you dictate into a ready document: meeting minutes, a task list, an email, a post, or your own templates: you can create and edit them in the web app, the desktop app and the mobile app (Android). They show up in all the apps.
A voice note with structure from the start
Dictate in the desktop or mobile app and, if AI text processing is turned on in settings, the note comes out already laid out in headings and lists, in Markdown. Without it the note is saved as the recognised text.
About voice notes →Meeting minutes from a recording
Upload a meeting recording in the web app (“Transcribe a recording”), in the desktop app (“Transcribe”) or in the mobile app (“Transcribe a file”): Voisee returns the text with “Speaker 1”, “Speaker 2” labels and so on. You choose the speech language: Russian or another language from the list, and splitting by speaker works with any of them. The finished transcript is kept for 7 days and is visible wherever you are signed in: you can close the page and come back for the result. Recordings up to 500 MB and 3 hours.
Transcribe a recording →Dictation without a connection
Lost the internet? Keep dictating. The desktop and Android apps warn you that there is no connection, save the recording and send it for transcription on their own once you are back online. The text appears in your history.
How text cleanup tidies dictation
Real output on dictation with filler words and a self-correction. The filler is gone, “Wednesday, uh, no, Thursday” became “Thursday”, and Docker and JWT are written the usual way. The sample is Russian speech.
Raw dictation
ну вот давай встретимся в среду э-э нет в четверг в три часа как бы обсудим докер и jwt After cleanup
Click “Clean up text” and the cleaned text will appear here.
Russian is fast, other languages work too
You pick the recognition mode in the app settings. Fast mode understands Russian speech only: it writes numbers as digits and common English terms in Latin script. Multilingual mode recognises English and other languages, but responds more slowly.
HDIPYouTube1С902ещёвсё+ text cleanup
Roadmap
What we are working on now. The order may change — tell us what you need.
- 01Planned
Video by link — translated into another language
Paste a link to a video and get its text and subtitles, translated into the language you need if you like. No need to save the file first, even for lectures, webinars and podcasts that run for hours.
- Links to YouTube, Rutube and VK Video
- Long recordings: lectures, webinars, podcasts several hours long
- Translation of text and subtitles into any language
- SRT and VTT subtitles with timecodes, straight in the language you need
Frequently asked questions
Still have questions? Write to admin@nizamov.school.
How well does Voisee recognise speech?
It works best with calm speech in a quiet room: the text comes out with commas, periods and capital letters. Fast conversational speech and background noise lead to more mistakes. Filler words and slips of the tongue are removed by text cleanup, which you turn on in the app settings.
Which languages are supported?
Russian and other languages. Fast mode recognises Russian speech only, and common English words inside it, such as IP, YouTube and HD, come out in Latin script. For English speech and other languages there is multilingual mode: it is slower, and one minute of recording takes 2.5–3.5 seconds. The desktop and mobile apps offer four options: “Russian — fast”, “Russian — slower” (multilingual mode with the language set in advance), “Other languages” with a language picker, and “Auto”, which detects the language for you.
How fast does the text arrive?
In fast mode, one minute of recording is recognised in about 0.1 seconds; in multilingual mode, in 2.5–3.5 seconds (measured on the production server: 106 minutes of Russian speech in 11 s in fast mode, 14 minutes of English in 36 s in multilingual mode). On top of that comes sending the recording to the server and waiting in the task queue, so the time until you get the text also depends on the recording length and your connection speed.
How are speakers separated?
A separate model listens to the recording and groups the lines by voice: “Speaker 1”, “Speaker 2” and so on. It works for any recognition language. On noisy recordings, when people interrupt each other or talk at once, voices are told apart less well: labels can be wrong, and one person can get two labels. You turn it on with a checkbox on the “Transcribe a recording” page.
Sign up and get 60 min of dictation every week at no cost
No bank card required.