About FreeTranscribe

FreeTranscribe turns video and audio into text, timestamps and subtitles, for free, without an account, and without uploading your file. It's an independent project, launched in July 2026, and it's now used by thousands of people in more than 150 countries for meetings, lectures, interviews, podcasts, voice notes and social video.

Why it exists

Most "free" transcription sites have a catch: you have to sign up, the free tier stops after a few minutes, the export is locked, or your recording is uploaded to a server you know nothing about. Browsers are now capable of running a modern speech-recognition model directly, so FreeTranscribe does exactly that. Because there's no server doing the work for each file, there's no per-minute cost to pass on, and your recording never has to leave your device.

How it works

• The model: transcription uses Whisper, the open-source speech-recognition model released by OpenAI. The Fast setting uses Whisper tiny and Accurate uses Whisper base. Both understand about 100 languages and can translate speech into English.
• In your browser: the model runs on your own device through Transformers.js and ONNX Runtime (WebAssembly). Formats the browser can't decode by itself, such as MKV or AVI, go through a WebAssembly build of FFmpeg, also locally.
• Downloads, not uploads: the first time you transcribe, your browser downloads the model files (from Hugging Face) and the speech engine (from jsDelivr) and caches them. After that, the only thing that moves is your audio, from your file into the model on your device. You can check this yourself.
• Outputs: plain text, text with timestamps, SRT and VTT subtitles (in standard, single-line and short social styles), Word documents and PDF.

How it's funded

FreeTranscribe is supported by advertising. Ads never see your files or transcripts. There's no paid plan and no feature held back behind one. Our privacy policy explains the advertising and analytics the site uses and how to opt out.

What it doesn't do (yet)

Being honest about limits saves you time:

• It doesn't label who is speaking. Transcripts of conversations need speaker names added by hand.
• Accuracy depends on the recording. Clear speech transcribes very well; music, crosstalk and phone-quality audio make more mistakes. Our accuracy test has the numbers.
• Very long recordings are much faster on a laptop or desktop than on a phone, because your device does the work.
• It doesn't download videos from YouTube or other sites. It works with files you have. Creators can download their own videos from YouTube Studio.

Built with

FreeTranscribe stands on open-source work: OpenAI's Whisper, Hugging Face's Transformers.js and model hosting, Microsoft's ONNX Runtime, FFmpeg and ffmpeg.wasm, and the Astro web framework. Thank you to everyone who builds and maintains them.

Learn more

• Guides: subtitles for YouTube, SRT vs VTT, transcribing on a phone, and more
• Privacy policy
• Terms of use