An interactive walkthrough
From sound
to knowledge
Nine scenes following one recording all the way through Scriptorium — from a pressure wave nobody can search, to a line of text you can quote. Scroll to move through it.
The dialogue in this walkthrough is sample text, and the waveforms are drawn. Nothing on this page is a recording being transcribed live.
Scene 01
Every meeting starts as noise.
A recording is a pressure wave written down. It holds everything that was said and none of it is findable — you cannot search it, quote it, or hand it to somebody who missed the room. Move your cursor: that is all the information there is so far.
Scene 02
Then it becomes a file on somebody's laptop.
Drop it in as it is. Video or audio, 11 video containers and 11 audio ones, no conversion step first. The picture is discarded immediately — only the audio track matters, which is why a 4K recording and a phone recording of the same talk cost the same to transcribe.
Scene 03
Inside, it comes apart.
The audio is decoded once and handed to the engine — 6 models, 33 languages plus auto-detect, running on the machine that holds the file. Speech becomes timed text; silence, throat-clearing and the half-sentence nobody finished do not survive the trip.
Scene 04
If more than one person was talking, that survives too.
Each voice is separated from the mix and given its own lane before a word of it is written down — up to eight speakers, worked out from the sound of each voice rather than guessed from turn-taking. Good at hearing that the speaker changed; imperfect at deciding exactly how many people there are when voices are similar or crosstalk is heavy.
Scene 05
It reassembles as something you can read.
Every line carries a start time, an end time and the speaker who said it. Word-level timing means a click on any word takes you to the second it was said. Click a line that is wrong and fix it; the correction is what gets indexed from then on.
Scene 06
Then you can find one sentence in all of it.
Full-text search across every transcript you own, not just the one open in front of you. It searches your corrections rather than the engine's first guess, which is the difference between a search box and a search that works.
Scene 07
A pile of recordings becomes a body of knowledge.
Once the words are indexed, the recordings stop being a folder of files you will never open again. Who said it, when, in which meeting, in which words — the things that were only ever in somebody's memory become things you can look up.
Scene 08
And none of it had to leave.
Self-hosted, the engine runs beside the files it processes: audio, transcripts, search and speaker analysis stay on infrastructure you control. The hosted workspace on this site is the other choice, plainly labelled — files there are processed on this server and free-preview files are deleted after 24 hours.
Scene 09
One recording. One line you can read.
That is the whole product. Everything else — the speaker labels, the word timings, the seven export formats, the editing, the search — exists to make that line true and useful.
Now the short version
Your meetings are
already the record.
They are just not readable yet. That takes one file and a few minutes.
Free for one file up to 15 minutes, no account · No card, no install