Audio to text converter
Wavetotext is a free audio to text converter with speaker labels. It runs OpenAI's open source Whisper model in your browser, so your files never leave your device.
Pick files or press Record to start.
Transcript
Saved on this device
Runs on your device. Models download the first time, then your browser keeps them. PDF opens your print window: choose Save as PDF.
How to convert audio to text
- Choose one or more files, drop them on the box, or press Record to use your microphone.
- Pick the language spoken, or leave it on Detect automatically. Turn on Speaker labels for interviews and meetings.
- Press Transcribe. Files run one after another, and text shows up piece by piece.
- Play the audio, fix any mistakes, rename speakers, then download Word, PDF, TXT, SRT or VTT.
What you get
- Free, no account No sign up, no paid plan and no minute limits from us. Your device's memory is the only limit.
- No upload Whisper runs on your device. Your audio is never sent to a server.
- Many files at once Drop a whole folder of recordings. They run one after another.
- Speaker labels See who said what, then rename Speaker 1 and 2 to real names.
- 99 languages Pick one or let Whisper detect it. Translate to English too.
- Graphics card speed On browsers with WebGPU, your graphics card runs the model. If it turns out slow, the processor takes over.
- Edit with the audio Click any line to hear it, then fix the words right there.
- Word, PDF, subtitles Download DOCX, PDF, TXT, SRT or VTT, with or without timestamps.
Which quality to pick
| Quality | Model | Download | Use it for |
|---|---|---|---|
| Fast | Whisper tiny | About 45 MB (80 MB on graphics card) | Clear English speech, older phones, a quick look |
| Balanced | Whisper base | About 80 MB (150 MB on graphics card) | Most recordings on a laptop or recent phone |
| Best | Whisper small | About 250 MB (490 MB on graphics card) | Accents, background noise, languages other than English, translation |
Each model downloads once. After that it loads from your browser's storage. Speaker labels add about 33 MB the first time.
Made for
Convert any format
Transcribe audio in your language
Questions
Is it really free?
Yes. There is no account, no sign up and no paid plan. The work happens on your own device, so there is no server bill to pass on. The only limit is your device's memory: long files need more of it.
Is my audio uploaded?
No. Your browser opens the file and runs the speech model on your device. Only the models are downloaded, from Hugging Face. Your audio and transcripts are never sent anywhere.
What files can I convert?
MP3, WAV, M4A, OGG, Opus, FLAC, AAC, MP4 and WebM open straight away in most browsers. AVI, MOV, MKV and other videos your browser can't open go through a built-in converter, which downloads about 10 MB the first time.
Can I transcribe many files at once?
Yes. Pick or drop as many files as you want. They join a list and run one after another, and you can open each transcript as soon as it starts.
Can it tell speakers apart?
Yes. Set Speaker labels to Detect speakers, or to the number of people if you know it. Each part of the transcript gets a Speaker 1, Speaker 2 label that you can rename. Similar voices can get mixed up, so check before you share.
Which languages does it understand?
All 99 languages Whisper was trained on. Pick the language spoken or let it detect it. Results are best in widely spoken languages and weaker in languages with less training data.
Does it work on my phone?
Yes, in recent mobile browsers. Phones are slower than laptops, so Fast or Balanced quality works best. Keep the page open and the screen on: phones often pause pages in the background.
How long does it take?
It depends on your device, the file length and the quality you pick. A laptop is much faster than a phone, and using your graphics card is faster again. You can read the first parts while the rest is still working.
Is there a file size limit?
There is no set limit, but the whole file is loaded into memory. Long recordings need more memory, so on phones, shorter files and Fast quality are safer. In our tests, a 1 hour recording finished on a computer using the graphics card, and a 2 hour file loaded and started.
How good is the transcript?
Clear speech in a quiet room comes out well. Accents, crosstalk, music and noise cause mistakes, and so do names and rare words. Always read it through before you use it.
Does it work with noisy audio?
Some background noise is fine. Loud music, crowds and wind cause mistakes. Pick Best quality for noisy recordings, and record close to the speaker when you can.
Can I edit the transcript?
Yes. Click any line to fix it, and click its time to hear that part. Your changes go into every download.
Which formats can I download?
Word (DOCX), PDF, plain text (TXT), and SRT or VTT subtitles. Timestamps and speaker names are included when they're switched on.
Where are my transcripts saved?
In this browser only, under Saved on this device. They stay until you delete them or clear the site data. The audio itself is not kept.
Do I need to install anything?
No. It runs in your browser. The speech model downloads the first time you transcribe and then stays in your browser's storage.
Is it better than typing it up by hand?
It's much faster and free, but a careful human typist is still more accurate on hard audio. A good routine: transcribe here, then read it through while the audio plays and fix what's wrong.