* Most VRS that I’ve had any experience with, usually require the mic to be pretty close to the speakers mouth, in order to separate the voice from any background noise.
But Lawyers/Barristers and Witness tend not to stand anywhere near the mics in court.
As a result, the mic gain on the mixer usually has to be cranked fairly high, which results in the mic’s picking up more background noise, thus reducing the effectiveness of the recognition software.

* Many people hate what their voice sounds like when they hear it amplified or recorded.
This is because what they normally hear when they speak also contains the acoustics of their own head (ie: bone conduction).
As a result, when they listen to the recording or amplification of the sound that comes out of their mouth, it sounds alien to what they think they sound like.
In this circumstance, a witness amplified to the court will quite often back away from the mic if they hear themselves through the PA

* The VRS’s I’ve seen are designed to work on one voice at a time, and have to be trained to recognize that voice. Which is a real bummer when you put all the work into training the system, and then you get a cold.

* also, As they’re designed to work on one voice at a time, for a VRS to have a hope of working effectively in a court, you’d need to record each mic on it’s own channel – effectively running 5 to 10 VRS’s per court, depending on the number of mics in the court room.

* And which voice gets recognized when two people start talking at once?

*personnel in court change from day to day, so for a VRS to be effective, it would have to be able to automatically recognize ANY voice, WITHOUT training, in room acoustic’s that range from ‘nice & dry’, to rooms with the acoustic properties of a great big bathroom (as many courtrooms in heritage buildings have)

and so-on, and so-on, and so-on

I’ll stop bitching now!