The use of Speechmatics to obtain a printed text is a great saving on employing a stenographer to do the work. I have experimented with different files to obtain the most correct printed text from the voice. Speechmatics themselves say that a good level is essential for the system to work. Strangely the results do vary enormously.. One male voice can sound very similar to another but the resulting text print out can be totally different. By way of experiment I recently tried a female archive voice and only used one side of the recording. The result I think was slightly better in that A) The female voice works better with Speechmatics anyway and B) The use of one side of the recording produced near perfect results especially when there were few technical terms and/or proper names. It is always worth an experiment to do a short before getting the whole interview processed. One further difficulty with the Speechmatics system is that it does not cope well where an interview has multiple voices who keep interrupting with asides.This could be well kept in mind for future interviewing,. Interviews should be as clean as possible with one voice which will assist the Speechmatics process a lot. To alter the transcription of a Speechmatics version does take an amount of time however. Personally I try to get all the technical terms correct along with names and worry less about the absolute grammar of the transcription.I believe that a subscriber will understand what is being said.