MotionShine01 / Start
Allow the microphone
Press Start and allow the microphone. The meter moves when it hears you.
Free tool
Voice typing, just talk.
Press Start, speak, and pause for a moment: each sentence is written out with punctuation and typed where your cursor is. Fix a word by typing, say "new paragraph" to start one, then copy the text or save it. The AI speech model runs in this tab, so your voice is never sent to a server. No account, no time limit.
Say
- "New line" for a line break
- "New paragraph" for a blank line
- Commas, full stops and question marks are added from how you speak
Tips
Pause briefly after each sentence or two; that is when it is written out. A headset or a quiet room gives the best result. The first time, the page downloads the speech model, about 130 to 160 MB, and keeps it for next time.
English only for now.
Your voice is turned into text in your browser. The text is kept in this browser only.
How it works
Speak, pause, it is typed.
02 / Talk
Speak naturally
A voice detector hears when you start and stop. Each stretch of speech is written out by an AI model on your computer.
03 / Use it
Edit, copy, save
Type to fix anything, then copy the text or save a .txt file. It stays here if you close the tab.
Questions
Good to know.
Is my voice uploaded?
No. The sound from your microphone is turned into text by AI models that run in this tab: Silero VAD hears when you speak, and Moonshine, an open speech model from Useful Sensors under the MIT licence, writes it out. Both are downloaded once from Hugging Face and kept by your browser.
Which languages work?
English. Moonshine's models for other languages are under a licence that does not allow commercial use, so they are not used here. To transcribe a recording in another language, use audio to text.
Why does the text appear after I pause?
Each stretch of speech is written out when you stop talking for about two thirds of a second, or after 20 seconds of non-stop talking. Whole sentences come out more accurately than single words.
Can I type a file instead?
Yes, with audio to text, which writes out recordings and videos, with times and subtitles.
Does it work on my phone?
It needs a browser that can run the model and record from the microphone, such as a current Chrome or Safari, and enough memory. On a computer it is quicker.