All tools

Audio to Text

Transcribe an MP3, WAV or voice memo to text and subtitles with Whisper — in your browser, nothing uploaded.

Select an audio or video file

MP3, WAV, M4A, MP4, MOV, WebM — transcribed in your browser with Whisper

Click to choose, or drop a file here

About transcribing audio in the browser

This transcribes an audio recording — a lecture, an interview, a voice memo, a podcast — into text and optional SRT/VTT subtitles, without sending the file anywhere. OpenAI's Whisper model runs locally in your browser through WebAssembly.

The model weights download once (~40 MB Tiny, ~80 MB Base, quantised) and are then cached by the browser, so subsequent runs start immediately and work offline. Base is more accurate; Tiny is faster. Whisper handles punctuation and capitalisation, auto-detects the spoken language, and can translate other languages into English on request. It does not do speaker labels or paragraph breaks.

Transcription runs at roughly 1–5× real time on WebAssembly, so a long recording takes a while — trimming to the section you need, or using Tiny, speeds things up. Accuracy drops on poor audio, heavy accents and crosstalk, so treat the output as a strong first draft to proofread.

Supported files

Accepts MP3, WAV, M4A, AAC, FLAC, OGG and the audio from most video files. Exports .txt, .srt and .vtt.

Is my file uploaded?

No. The model runs in your browser with WebAssembly — the recording stays on your device.

Common problems

It's taking a long time
That's expected for long files on WASM. Use the Tiny model, or split the audio and transcribe the parts you need.
Wrong language detected
Set the language explicitly instead of leaving it on Auto-detect.
The download bar is stuck
The model files are tens of megabytes — give it a moment on a slow connection. It only happens once per model size.

How to Audio to Text

  1. 1

    Add your audio

    Drop in an MP3, WAV, M4A or voice memo — it's transcribed in your browser.

  2. 2

    Pick a model and language

    Base is more accurate, Tiny is faster. Auto-detect handles the language, or choose one. The Whisper model downloads once then works offline.

  3. 3

    Transcribe and export

    Click Transcribe, then copy or download the result as text, SRT or VTT.

Related PDF tools

Audio to Text — frequently asked questions

Is my audio uploaded?+

No. Whisper runs locally in your browser with WebAssembly — the file stays on your device.

How long can the audio be?+

There's no hard limit, but it's processed in memory and transcription runs at roughly 1–5× real time on WASM, so a one-hour recording can take a while. Tiny is faster than Base.

Does it add punctuation and paragraphs?+

Whisper adds punctuation and capitalisation. It doesn't do speaker labels (diarisation) or paragraph structure.

Can it translate?+

Yes — tick "Translate to English" and non-English speech is transcribed straight into English.

Do you upload my files to a server?+

No — every tool here processes your files entirely on your device, and nothing is uploaded to a server. The only exception is the AI tools, which send the extracted text (not the file) to Claude to do their work.

Is there a file size limit?+

There's no artificial limit we impose, but since processing happens in your browser, very large files (500+ MB or 1,000+ pages) are limited by your device's memory rather than by us.