Free voice keyboard

Free voice typing in English on Windows

Press a hotkey, talk, and the words land wherever your cursor is: an email, a chat, a code comment, a prompt for an AI. Here is how VoiceKeyboard handles English and which engine fits which job.

A vintage brass microphone next to a green dot-matrix display that reads

Three steps

  1. Put the cursor where the text should go. Any window works: Word, Gmail in a browser, Slack, Notepad, VS Code.
  2. Press Ctrl+1 and speak. The POWER tube warms up while VoiceKeyboard is listening.
  3. Press Ctrl+1 again to stop. With online engines the text is typed as you talk; with most offline models it appears after each phrase.

You can move the hotkey to another combination in the app if Ctrl+1 is taken.

Which engine to use for English

Every engine in VoiceKeyboard understands English, except the two models made only for Russian, so the choice comes down to what matters to you: speed, privacy or a mix of languages.

EngineInternetText appearsPlanPick it when
Sonioxyeswhile you speakProyou want the best accuracy, long meetings, or you switch between English and another language mid-sentence
Standard (Google)yeswhile you speakFreeyou want live text for free and you speak one language at a time
Parakeet v3noafter each phraseFreeyou want offline dictation with punctuation and a good speed on an ordinary PC
Whispernoafter each phraseFreeyou dictate in several languages offline; bigger sizes are more accurate but slower
Vosknowhile you speakFreethe computer is old or slow; the English model is small

Offline models download once, inside the app, and after that your audio never leaves the computer. Online engines send the audio to the provider only to transcribe it; VoiceKeyboard does not keep recordings.

The RECOGNITION panel: a big slide switch between ONLINE and LOCAL above the green model screen, the credits gauge and the orange Quality and Speed lamps.
The ONLINE / LOCAL switch picks the family of engines; the lamps show how each one rates on quality and speed.

Accents and other languages

If English is not your only language, two engines help. Soniox listens for the language itself and can switch inside a single sentence, so a German or Spanish phrase in the middle of an English email stays readable. Whisper on the AUTO key works out the language of each recording, so dictate the other-language part as its own recording. The Standard engine hears exactly one language, the one whose key is lit.

The language keys sit on the RECOGNITION panel. Press + to choose the languages you actually use, out of 33, and each gets its own key. VoiceKeyboard remembers the language separately for every engine, so switching engines does not reset it.

Free or Pro for English

Free

  • 2,000 words a day with any engine, online or offline. That is several long emails or about four pages of notes.
  • All offline models, including Parakeet, Whisper and Vosk.
  • One audio or video file and one video link a day, up to 60 minutes each.
  • The catch: no Soniox, and the daily limit stops dictation until midnight.

Pro

  • No daily word limit on the Standard engine and offline models.
  • Soniox: 1,500 minutes a month, live text, language detection, recordings as long as a meeting.
  • Files and links without limits, several at once, with timestamps.
  • The catch: Soniox minutes do not carry over to the next month. When they run out, dictation keeps going on the Standard engine.

Pro costs $14.99 a month, or $9.99 a month on a yearly plan, and comes with a 14-day refund. It is the same app; a license code switches Pro on.

Small things that save time

  • Dictating prompts for ChatGPT or Claude is where voice pays off most: you say in twenty seconds what takes two minutes to type.
  • The counters on the POWER panel show how many words you dictated and roughly how much typing time that saved.
  • If someone is reading over your shoulder, switch the live transcript screen off. The text still goes to your window.
The VoiceKeyboard window in English: a glowing vacuum tube on a brass POWER panel, the RECOGNITION panel with the model screen and language keys, and the live transcript screen.