Transcribing a deposition recorded on one microphone

Whether Pincite Audio is any use on a deposition is a question about the microphone rather than about the proceeding. The short answer is that a room recorded the ordinary way comes back with its speakers separated by a machine guess rather than by the recording, and that adding microphones is not the reliable fix it looks like.

One microphone in the room

A room recorded on a single microphone is one mixed signal. Nothing in the file says who spoke, so the recogniser groups the lines by voice and the transcript shows those groups as separate speaker tracks — an estimate, marked as one. How many people spoke is not established, and no track carries a name until someone at your firm puts one there. Where the recording keeps speakers on separate channels instead, that separation is a property of the file and nothing is estimated at all. The speaker guide explains the mechanism, and the answer is the same for a deposition as for a meeting or a panel.

Why more microphones may not help

The obvious fix is to record each participant separately and hand over a multi-track file. It is worth knowing, before you rig a room that way, that the product does not take separation on trust.

It measures it, and what it measures is overlap in time: a multi-track recording is treated as separated only where the tracks are mostly not carrying the same speech at the same moment. A two-party call recorded to two tracks often looks like that, because the parties mostly take turns — though not always, and a call with heavy talking-over scores low on the same measure. Microphones open around one table rarely look like it at all: each one picks up everyone who speaks, to some degree, and that overlap is exactly what the measurement is reading. A file the product cannot confidently call separated is routed to the single-speaker path.

That direction is deliberate, and the reasoning is worth stating because it decides the behaviour. Treating a mixed recording as separated would attribute every word to both tracks, so each one would appear to contain the whole conversation. Failing the other way costs you speaker labels you did not have anyway. The thresholds behind the decision are uncalibrated, and the choice of which way to fail was made accordingly.

So a microphone array is not a reliable route to a speaker-labelled deposition transcript here, and a firm planning capture around one is worth telling before the day rather than after the upload.

The deposition setting, and what it is not

There is a deposition option at upload. The kind you declare decides how the finished document is cited, among other things, and the format page describes the two citation styles.

It adds no speaker separation. Choosing it does not turn a one-microphone recording into a transcript that names speakers, which is worth saying plainly, because a setting carrying a proceeding's name reads like a mode.

What this site does not publish

No accuracy figure appears anywhere on this site, for room audio or for anything else. That is a policy rather than a gap in this page, and it is why this one describes what the product does with a recording rather than how well it does it.