I must admit that I was a bit sceptical when I first read the instructions for Voice Master - voice production through the TV speaker under software control? Never!

Well, it does. The Unit is a small silver box with one or two knobs to twiddle, a small LED, and wires from the box to control port 2 and the headset. The box can also be connected to the audio/video socket on the back of the machine so that you can hear the sound through the headset speaker.

The headset is similar to those 'Walkman' type headphones, except that only one ear is covered. There is also a short tube coming from the earpiece to a small, foam-covered, microphone that sits comfortably in front of your mouth.

All the leads are of a reasonable length and I didn't have any problems with snagging cables or having to rest my head on the keyboard. Two other cables are supplied to connect the unit to either the A/V socket as previously mentioned, or onto a standard Commodore monitor cable. Luckily, my decidedly non-standard cable and monitor fitted nicely.

When everything is plugged in, you can load up the software and start calibrating the device. Calibration involves shoving a small (supplied) piece of metal into a slot, tweaking this with the software provided, until the right number flashes on the screen. With this sorted out, you can go onto the rest of the goodies supplied on this disk.

There are three things you can do with this unit - record and playback speech, speech recognition, and other things. The recording and playable system uses an extension to the resident Basic interpreter to work. Commands such as LEARN, SPEAK, RATE and SPEED, govern the learning, talking, sampling rate, and speed that the words are spoken respectively. The higher the sampling rate is, the higher the quality of the reproduced speech. Arguments after Learn and Speak refer to a set of 64 numbers that are the recorded or spoken phrases. Thus Learn 1 will record a word or phrase and Speak 1 will say the recorded word.

There are various other keyboards that deal with upkeep of the word lists and storage of words on disk and tape.

The speech recognition system is also driven via basic keywords. RECOG will activate a template of pre-recorded words and compare them to whatever is being spoken. Both the recognition system and the digitisation system can be in memory at one time for a fully flexible voice system.

The best thing is that to play back speech you don't need to have the box connected, it all comes out of the normal speaker using the SID chip. For recognition, however, you do need the little box to digitised the incoming speech.

There are various other things that you can do with the box and there is software to let you do it. The voice harp has featured heavily in Anirog's advertising. This is a system where you can whistle or hum a tune and notes are created for you. But don't expect to ust whistle a merry tune and have it come out in the correct musical notation, it does need some work and skill to use; even still, it is an excellent piece of software and once you get used to it you can compose simple tunes.

Software for a speaking clock and a talking calculator is also on the disk; they are not very useful, but they answer the "yes, but what does it do, dear?" question to astounded onlookers.

The manual is crudely presented, but the information provided in it is of an extremely high standard. I much prefer this type of manual to the highly glossy, but utterly useless type so often found in an otherwise excellent piece of equipment.

Nobody should have any difficulty with understanding the system, either for the occasional dabbler or the expert trying to create a HAL-type talking and listening computer. I can safely say that this product surprised me thoroughly, and I am very pleased with it. The package is totally comprehensive and you can have your computer shouting at you in no time.

If you want to experiment with speech and speech recognition then rush out and buy one.