Voice to Text
betaThis free voice-to-text tool turns speech into editable text, timestamps and subtitles right in your browser. Whisper models transcribe files and recordings accurately in English, Indonesian and many other languages, and a 17 MB Fast engine powers live dictation as you speak in English, German, French, Spanish, Italian, Dutch and Polish — even on phones. Your audio is processed on your device and never uploaded.
See words appear as you speak (uses the Fast engine).
How to use Voice to Text
- 1 Click Record to capture from your mic, tick Live dictation to see words as you speak, or drop an audio or video file (mp3, wav, m4a, mp4…).
- 2 Choose the spoken language and a model. Multilingual (Whisper) is the default for files and recordings; pick Fast for quicker results in English, German, French, Spanish, Italian, Dutch or Polish.
- 3 Click Transcribe (or stop live dictation); the model downloads once on first use, then runs on your device.
- 4 Switch between the Text, Timestamped and Subtitles tabs, edit the text, and download .txt, .srt or .vtt.
Frequently asked questions
Is my audio uploaded to a server?
No. Transcription runs entirely in your browser with on-device models. Your recording or file never leaves your device, so private conversations stay private.
Which model should I use?
Multilingual (Whisper) is the default for files and recordings and works for English, Indonesian and many other languages. Multilingual · Better is the most accurate but much slower. Fast (17 MB) is the quickest option for English, German, French, Spanish, Italian, Dutch and Polish, and it always powers live dictation.
How does live dictation work?
Tick Live dictation and press the button: your speech is transcribed about once a second on your device. Confirmed words appear solid and the latest guess is shown faded until it settles. When you stop, you get the full transcript and subtitles, and the recording is kept so you can replay it.
Can I export subtitles?
Yes. Open the Subtitles tab and download SRT or VTT files with timestamps, or grab plain text (.txt) or a timestamped version instead.
Why does the first run take a while?
The model is downloaded and cached on first use (17 MB for Fast; Whisper models are much larger). After that it loads from your browser's cache — even offline — and transcription runs locally.