Transcribe audio and video to text, without uploading the file
Meetings, lectures, interviews, podcasts, voice memos, videos that need captions: drop the file, get the text and an .srt subtitle file. The AI (Whisper) runs in your browser.
The first run downloads the model (about 80 MB). After that it stays cached.
How it works
- The Whisper speech recognition model downloads to your browser (once).
- Your file is read on your device and split into 30-second parts, cut at pauses.
- Text appears as it goes, with a timestamp for each sentence, ready as .txt, .srt or .vtt.
What people use it for
Meeting notes
Record the call, transcribe it, keep the text. Nothing goes to a third-party server.
Video captions
Shorts, Reels, TikTok, YouTube: a timed .srt to drop into your editor.
Lectures and interviews
Turn a recording into notes you can read, search and quote.
FAQ
Is it really free?
Yes: up to 10 minutes per file with the Fast model, no account, no limit on the number of files. Pro ($7/month) goes up to 2 hours per file and adds the Accurate model.
Are my files uploaded anywhere?
No. The AI model downloads to your browser once, then all the work happens on your computer or phone. You can check: once the model is loaded, the tool keeps working with your internet turned off.
Which formats work?
Anything your browser can play: MP3, M4A, WAV, OGG, FLAC, WEBM, and the audio track of MP4 or MOV videos. You can also record straight from your microphone.
How long does it take?
It depends on your device. On a recent computer with WebGPU, the Fast model is often faster than real time. On an older laptop or a phone, expect several minutes for 10 minutes of audio. The time left is shown while it runs.
How do I make subtitles for a video?
Drop the video and download the .srt file. It imports into CapCut, Premiere Pro, DaVinci Resolve, YouTube and VLC. For subtitles in another language, open the .srt in the Translate tool.
Which languages?
Whisper recognizes about 100 languages, including English, Spanish, French, German, Portuguese, Italian and Arabic. Picking the spoken language instead of auto-detect improves accuracy.