Writing plays using speech recognition software ought to be even more difficult than writing novels that way. In this guest post, playwright Geoff Bamber explains how he does it…
I’d like to say that my use of voice recognition software was due my being my being up-to-speed/ahead of the curve and indeed technologically savvy but no – it was and, to an extent, still is mainly down to pure laziness.
Back in the days when I thought I was a novelist, the notion of typing out a hundred thousand words or more with two slow fingers on a keyboard attached to a steam-driven computer was a bit daunting. As I never found stream-of-consciousness babble a problem, the idea of not having to physically write it all down was very appealing. Short of morphing into Barbara Cartland, employing a secretary and dictating to her/him from a reclining position on a chaise longue in a room that looked like the inside of a marshmallow, voice-to-text was as good as it got.
I was enticed to start with IBM Via Voice. If I remember correctly, the publicity material showed a sharp executive leaning back his chair with his feet up on his desk dictating a business letter which he would not need to check or proof read at all before it was fit to send. As these were the days when people still sent letters with stamps on them he would probably have to sign it himself but the general tone of the marketing was that the communication would climb into an envelope of its own accord and that was the last he would see of it, thus allowing him to take the afternoon off for a round of golf.
Unfortunately real life isn’t like that. The software had to be ‘trained’ to be better able to recognise the speaker’s intonation though a heavy regional accent would always be a problem. Thus my first requirement was to tone down my northern vowels (best achieved by taking the flat cap off and making sure the whippet was in the other room) and speaking slowly in standard English.
I must say my early experiences were not particularly successful but the software has got better and so have I. I have used two or three other programmes over the years – currently Dragon Naturally Speaking. Like the original ViaVoice, it does ‘learn’ to follow my dictation but is by no means foolproof. It seems happier with American pronunciation and thus has trouble with seemingly simple words like ‘ladder’ and ‘daughter’.
There is a tendency to over-compensate to the point where I speak unnaturally slowly and without any expression and end up sounding like Stephen Hawking’s voice synthesiser. In actual fact ‘normal’ speech certainly fares no worse. Even so when writing scripts, my first draft is always written on paper with a fountain pen(!) before being dictated into the software. I only normally dictate actual dialogue. The speaker and any stage directions are typed in by hand afterwards while the various faux pas of the voice recognition system are hopefully being picked out and corrected.
Dictating straight from imagination to screen leaves me dangerously open to reading the material afterwards and having no idea what I said as the software has concocted something that doesn’t make any sense at all. Things tend to go haywire when I speak too quickly and when the microphone is too far away to pick up dictation clearly. (I use a call centre-style headset.)
A good typist may well probably not find the process any quicker than typing the whole script in manually but it’s a lot less wear and tear on ageing fingers, shoulders and neck and a similarly ageing keyboard.
Over the years my typing speed has got faster but not a lot more accurate. For me voice recognition, even with me speaking slowly, sets out text at twice the speed that I can type and makes fewer errors. At ‘normal’ speed (i.e. how the actors might be expected to deliver the lines) the time can be halved again with only a slightly higher error count.
The major drawback is being interrupted in mid-flow, either by the dog barking, other members of the family coming into the room or me answering the phone while neglecting to switch the mike off.
I would point out that my software programme, though I am quite happy with it, is from the cheaper end of the market and that more sophisticated and presumably more accurate versions are available.
I’m even lazier now than I was when I started using voice recognition software so I’d be reluctant to abandon it and would recommend anyone to give it a try. Just work on that Californian accent and you can’t go wrong.
[The results, in the form of Geoff’s plays, can be explored here.]