Pick a recording and get the text right here in this browser tab. The page runs OpenAI's Whisper on your own computer, so the file stays with you. There is no account, no watermark and no limit on minutes.
Checking what your browser supports...
[00:00:05] Thanks for joining, let's start with the schedule.
[00:00:11] The first deadline is the end of next week.
[00:00:18] I'll send the notes to everyone this afternoon.
Would you rather talk than type? StarWhisper lets you use the same Whisper in any text box on your PC or Mac. Press a shortcut, speak, and the words land in your email, document or chat. Free for 2,000 words a week, no account.
Try it freeVoice notes. WhatsApp voice notes are Opus files, and this page reads them. Save the note as a file first: on WhatsApp Web, open the message menu and click Download. On a phone, use the share option and save it to your files.
Meetings and interviews. A recording from Zoom, Teams or a voice recorder app works as it is. The file never leaves your computer, which matters when the conversation is with a client or a patient. Whisper writes down what was said but not who said it, so you will need to add names yourself.
Lectures and podcasts. There is no limit on length. A long recording just takes longer, so leave the tab open and come back to it.
Video to text. Pick the MP4, MOV, MKV or WEBM file and the page reads only its sound. In Chrome and Edge a video of several GB works on a computer, because the file is read in small pieces. Save the result as SRT and you have a subtitle file.
The models cover the 99 languages Whisper was trained on, including Hindi, Spanish, Portuguese, Arabic, Japanese and Indonesian. The language is detected automatically. If you tick "Translate the result to English", speech in another language comes out as English text. It can't translate into other languages.
Most free transcription sites upload your recording to their servers first. This page doesn't. Your browser reads the audio and hands it to Whisper running in the same tab. To check, open your browser's developer tools, go to the Network tab and transcribe something. You'll see the model files coming in from huggingface.co the first time, and no upload going out.
The page does send two kinds of data, but neither includes your file or your words. If you accept cookies, our usual site analytics load as they do on every page. And the page sends a few anonymous counts to our server, without a cookie: when you pick a file (its rough size and how the page reads it), when a transcription finishes (the model, CPU or GPU, a rough length such as "1 to 5 minutes", and the speed), and when something fails (the step that failed). The code is open source on GitHub, so you can read exactly what it does.
| Model | Download size (CPU / GPU) | Good for |
|---|---|---|
| Tiny | 41 MB / 120 MB | Quick drafts of very clear speech (best in English), older computers and phones |
| Base | 77 MB / 206 MB | Most recordings with one or two people speaking clearly |
| Small | 249 MB / 586 MB | Accents, noisy rooms, and anything where accuracy matters more than speed |
That depends on your computer, because your computer does the work. With a graphics card through WebGPU, an hour of audio can take a fraction of an hour. On the processor alone, a long recording can take longer than the recording itself. Chrome or Edge on a recent computer is fastest. The note under the file box tells you whether you're using the graphics card or the processor. Phones have far less memory, so the page only offers Tiny and Base there, and short recordings work best.
This page is for files you already have. To talk instead of type, the StarWhisper desktop app runs the same Whisper on Windows and Apple Silicon Macs. You press a shortcut, speak, and the text appears in whatever program you're using. It works offline once it's set up. The free plan covers 2,000 dictated words a week with no account, and Pro adds file transcription in the app, without a browser's memory limits, plus the Medium and Large models. Download StarWhisper.
Yes. Your computer does the work, so we have no per-minute cost and no reason to cap you. There's no sign-up, and nothing is added to your text.
No. The file is read and transcribed inside your browser. The only downloads are the page, the transcription libraries from jsDelivr and the model files from Hugging Face.
MP3, WAV, M4A, AAC, OGG, Opus, FLAC and WEBM audio, and the audio track of MP4, MOV, MKV and WEBM video. Chrome and Edge read the most formats. If a file won't open, convert it to MP3 or WAV first.
In Chrome and Edge, a video of several GB works on a computer. Browsers that can't read a file in pieces stop at 1 GB on a computer and 200 MB on a phone. FLAC files stop there in every browser.
The page itself needs a connection to load, along with the model the first time. The transcribing happens on your device either way.
No. Whisper writes what was said, not who said it.
The Whisper Web page covers how the tool works under the hood, including the models and how long files are split.