Transcribe Audio to Text

Turn a recording into text with AI speech recognition that runs entirely in your browser. Your audio is never uploaded, which makes it suitable for meetings, interviews and personal voice notes.

Your file never leaves your device. How?

Your media is processed in your browser and is not uploaded to our servers for conversion. Page assets (including the converter engine) still load over the network.

Select or drop an audio or video file

MP3, M4A, WAV, OGG, MP4, MOV, WebM and more. Transcribed on your device.

Click, tap, or drop a file

How to transcribe a recording

Choose the recording

Drop or select an audio file (MP3, M4A, WAV, OGG) or a video (MP4, MOV, WebM).

Set the language

Pick the spoken language for better accuracy, or leave it on automatic detection.

Transcribe

The first time, the speech model (about 41 MB) downloads and is cached. Then transcription runs on your device.

Copy or download

Copy the text, or download it as plain text, timestamped text, SRT or VTT.

How private transcription works

The tool runs Whisper, OpenAI's open-source speech recognition model, inside your browser using WebAssembly. The model file is downloaded from this site the first time and stored in your browser's cache; your recording is decoded and transcribed locally and is never sent anywhere.

It uses the compact 'tiny' version of Whisper so it can run on ordinary laptops and phones. It handles clear speech well and adds punctuation, but it makes more mistakes than large cloud services with background noise, strong accents, overlapping speakers or specialist vocabulary. It doesn't label who is speaking. Always review the transcript before relying on it.

Features

Never uploaded

The recording is processed on your device, so confidential conversations stay confidential.

Free and unlimited

No account, minutes quota or subscription.

Four output formats

Readable paragraphs, text with timestamps, or SRT and VTT subtitle files.

Works offline after setup

Once the model has downloaded, transcription doesn't need an internet connection.

Frequently asked questions

Is my audio uploaded to a server?

No. Only the speech model is downloaded, once, from this site. Your recording is transcribed in your browser and never leaves your device.

How accurate is it?

Good for clear speech in a quiet setting. Accuracy drops with background noise, music, heavy accents, people talking over each other and technical terms or names. Choosing the spoken language helps.

How long does it take?

On a laptop, roughly 4–5 times faster than real time: an 82-second clip took about 15–21 seconds on a 2020 laptop once the model was loaded, so short voice notes take seconds. Phones are slower. Keep the tab open while it works.

How long a recording can I transcribe?

Up to 2 hours and 500 MB on a computer, or 1 hour and 200 MB on a phone, because the audio is decoded in memory. For a large video, convert it to M4A with the Audio Converter first; the audio alone is much smaller.

Which languages are supported?

Over 20 are listed, including English, Hindi, Spanish, French, German, Portuguese, Arabic, Japanese, Chinese, Bengali, Tamil, Telugu, Marathi and Urdu. English gives the most accurate results with this model.

Does it identify different speakers?

No. The transcript is one continuous text. Timestamped text makes it easier to find who said what.