Is my audio or video uploaded?

No. The file is decoded and transcribed entirely in your browser — with a self-hosted Whisper model and a local ffmpeg engine. Nothing is sent to a server, and it works offline after the one-time model download.

What languages does it handle?

This uses the small English Whisper model (whisper-tiny.en), so it's built for English speech. Other languages will produce poor results.

How accurate is it, and how fast?

The tiny model is fast and works on modest devices, but it's less accurate than larger cloud models — expect the odd mistake, especially with accents, noise, or overlapping speech. WebGPU-capable browsers run it much faster than the WASM fallback.

Can I get subtitles with timestamps?

Yes. Alongside the plain transcript you can download .srt or .vtt subtitle files with per-line timestamps, ready to drop into a video editor or player.

Is there a length limit?

There's no hard limit, but transcription runs on your device, so long files take a while — for the best experience, transcribe shorter clips or split long recordings.

Audio & Video Transcriber

Transcribe audio or video to text on-device with Whisper — get a transcript plus .srt/.vtt subtitles, nothing uploaded.

Processed 100% in your browser — nothing you enter here is ever uploaded.

Common use cases

Caption a video for social media

Generate an .srt/.vtt subtitle file for a Reel, Short, or YouTube video without any upload or subscription.

Turn a voice note or meeting into text

Get a written transcript of a recording so you can search, quote, or summarize it later.

Draft notes from a lecture or podcast

Transcribe a talk to text you can skim and edit, instead of re-listening to the whole thing.

How to use the Transcriber

Drop an audio or video file into the box, or click to browse. Everything runs on your device — the file is never uploaded.

The first time you use it, the Whisper speech model (~40MB) downloads once and is cached in your browser; after that it works offline.

Click "Transcribe". The audio is extracted locally with ffmpeg, then the model converts speech to text — you'll see progress as it goes.

Read the transcript, then Copy it, or download subtitles as .srt or .vtt (with timestamps) or plain .txt.

This uses the small English Whisper model, so it's tuned for clear English speech. Shorter clips transcribe fastest — browsers with WebGPU are noticeably quicker than the WASM fallback.