Audio to a transcript you can actually read

Questions

Does my file get uploaded?

The audio does, and this is the only VidClip tool where anything leaves your machine. Your browser strips the audio out locally and downsamples it to mono 16 kHz first, so what gets sent is speech only — no video, and a small fraction of the original size. It is not stored or logged on our side and is discarded once the transcript returns. Every other tool on this site still processes entirely in your browser.

Why can't it run locally like the other tools?

Speech recognition needs a model in the hundreds of megabytes. Shipping that into a web page would mean an enormous download before the first word was transcribed, and it would still be less accurate than a current server-side model. That trade-off was worth being upfront about rather than hiding.

What does 'organised' mean?

Raw speech-to-text is one unbroken block with no punctuation structure. This adds a title, a short summary, headed sections at natural topic shifts, clean paragraphs, and any explicit next steps stated in the recording. Your words are preserved — it formats, it does not rewrite or summarise the body.

How long can a recording be?

About 100 minutes per run. The audio is compressed to roughly 32 kbps before upload, so the 25 MB request ceiling works out at a long meeting or a full podcast episode. For anything longer, trim it into parts first and transcribe each.

Which formats can I export?

Markdown with the full structure, or plain text of the unformatted transcript. You can also copy the Markdown straight to your clipboard.

Is it accurate?

Good on clear speech, and noticeably worse on heavy background noise, overlapping speakers, or very quiet recordings. Naming the spoken language instead of leaving it on auto-detect helps on short or noisy clips. Always read it back before relying on it for anything that matters.