MP3, WAV, M4A, MP4, MOV, WEBM, OGG, OPUS. Stays on your device.
Choose a file to start. Nothing is uploaded.
- Whisper on your deviceOpen-source model, runs in your browser
- 0 uploadsYour audio never reaches a server
- TXT, SRT, VTTCopy or download subtitles
- No sign-upNo account needed
How to do it
WAV files work directly. Large uncompressed WAVs are fine; they are decoded in memory on your device.
How it works
- Choose a file. Your browser reads the audio track locally.
- The first time, the Whisper model downloads (40-80 MB) and is cached for next time.
- The audio is transcribed on your device in parts, so text appears as it goes.
- Copy the text or download TXT, SRT or VTT.
Speed and accuracy
| Model | One-time download | Best for |
|---|---|---|
| Fast (Whisper tiny) | about 40 MB | Clear English speech, quick drafts, older devices |
| Accurate (Whisper base) | about 80 MB | Most uses, other languages, accents |
Speed depends on your device. On a typical laptop, expect very roughly 2 to 8 times faster than real time with the Fast model and somewhat slower with Accurate; phones are slower.
Honest limits
- It does not label who is speaking.
- Music, overlapping voices, and background noise reduce accuracy.
- Names and specialist terms often need correcting.
- Lower-resource languages are less accurate than English or Spanish with small models.
Your files stay private
Most free transcription sites upload your recording to their servers. ScribeRuler runs the model inside your browser, so the audio never leaves your device. Only the model files are downloaded.
Related guides
- MP3 to text converter
- Transcribe MP3 files online, free and private
- M4A to text (iPhone recordings)
- MP4 video to text
- Video to text converter (free, no sign-up)
- OGG / OPUS audio to text
- WEBM to text
- Audio to SRT subtitles generator
Frequently asked questions
Is ScribeRuler free?
Yes. It is free with no sign-up and no minute limits, because transcription runs on your own device.
Is my audio uploaded?
No. The file is processed in your browser. Only the model files are downloaded.
Why is the first run slow?
The speech model downloads once (40-80 MB). After that it is cached by your browser.