Voice Typing — Online Dictation with Microphone
A voice typing tool that turns speech into text live as you speak. Press the microphone, start talking, and your words land on the page the moment you say them — say "comma" or "new paragraph" to punctuate without lifting a hand. It is built for hands-free writing: capturing a thought, drafting an email, taking notes in a meeting, or writing when typing is uncomfortable. Recognition uses your browser's built-in speech engine.
How to Type with Your Voice
Press the mic button and allow your browser to use the microphone when it asks.
Speak naturally and your words appear in real time; say "comma", "period" or "new paragraph" to punctuate as you go.
Switch on continuous recording and focus mode to dictate at length, and drop a bookmark whenever something matters.
Tidy the text by hand, then copy it or save it as TXT, SRT, VTT, JSON, PDF or DOCX.
Dictate with your voice and watch the words appear as you talk
Already have a recording? 📁 Transcribe an audio file
Your transcribed text will appear here.
Transcription
📖 History
☁ Word Cloud
Visual map of your most frequent words — the bigger the word, the more often you said it.
🔖 Bookmarks
⌨ Keyboard Shortcuts
Which model to pick: in the browser or on the server
The model downloads once, stays cached and runs on your device, so the time depends on it.
Time is minutes of work per minute of audio, the median of visitors’ runs; the first run also waits for the model to download. Names match the “Recognition Quality” list.
| Model | One-time download | Computer | Phone | When to pick it |
|---|---|---|---|---|
| FastWhisper Tiny | 39–73 MB | ≈ 0.3 | 0.2–0.3 | A draft of English speech on a weak device. In other languages it errs so often that the tool never offers it on its own. |
| AccurateWhisper Base | 73 MB | ≈ 0.6 | 0.3–0.9 | English speech; other languages when the device can’t keep up with “High accuracy”. |
| High accuracyWhisper Small | 238–462 MB | ≈ 1.4 | 2.4–6.4 | Languages other than English and Russian, when the device keeps up: in Spanish it makes 3.5 times fewer word errors than Base. |
| MaximumWhisper Large v3 Turbo | 1–1.5 GB | ≈ 3.1 (estimate) | 5–14 (estimate) | The hardest audio. Needs a graphics card; not offered on devices with 4 GB of memory or less, and the tool never switches it on by itself. |
| Russian — more accurateGigaAM v3 | 305 MB | ≈ 0.6 | ≈ 2.7 (estimate) | Russian speech: the tool picks it itself when the device can handle the download. Several times fewer word errors than Whisper; recognizes no other language. |
For a video, only the audio track is transcribed — it is dozens of times smaller than the video itself.