About TryScriptorium

TryScriptorium (the product is also called Scriptorium) is a web app that turns speech in audio and video files into editable, timestamped text. It supports 33 languages plus automatic detection, including Arabic, and exports TXT, TXT with timestamps, SRT, VTT, Markdown, JSON (word timings) and CSV. One file at a time, up to 15 minutes and 25 MB, 3 tries per day per network address, no sign-up and no card.

It is a web app at tryscriptorium.com. The short, quotable version of everything on this page, with the current limits and prices, is the fact sheet.

Who builds and runs it

TryScriptorium is built and maintained by one person, not by a company with a support department. The contact page reaches that person directly.

What it is

Scriptorium turns speech in video and audio into text you can read, correct, search and export. That is the whole product. Everything else it does exists to make the transcript more useful once you have it.

Where it came from

It started as a tool for transcribing meetings without inviting a third-party bot into the call, and grew into a general transcription studio: speaker labels, word-level timing, inline editing, full-text search across everything you have transcribed, and seven export formats.

If you would rather watch than read, From sound to knowledge follows one recording through the product, scene by scene.

That origin still shapes it. The engine runs as a single self-contained process with the speech models on the same machine, rather than calling out to somebody else's transcription API. Which means the same software can be run two ways, and both are real.

Two ways to run it

Hosted, here. You upload a file to this server, it is transcribed here, and your transcripts stay with your account until you delete them. That is what this site is, and the privacy page describes it plainly.

Self-hosted. The same software runs on your own hardware, where nothing is uploaded anywhere at all. For recordings covered by legal privilege, medical confidentiality or a contractual obligation, this is the honest answer, and no hosted service — including this one — can substitute for it.

What we will not do

We do not publish an accuracy percentage, because the only honest answer depends on your audio rather than on our software, and any number we printed would be measured on a recording flattering to us. We do not invent user counts, testimonials or awards. And the things the product cannot do — fetching a YouTube link, burning captions into video, an API — are listed on the homepage rather than left for you to discover after uploading something.

Found a problem, or want something it does not do? The contact page reaches the person who builds it.