From the time the Daleks trundled across our TV screens muttering "Exterminate, Exterminate" in their chillingly inhuman voice, I have been intrigued by speech synthesizers. At the time of writing, and to the best of my knowledge, there are six different units on the market to enable your micro to answer you back. Two others have recently lost their voice. I asked why, but the suppliers aren't talking.

The units on the market at the moment are Sweet Talker from Cheetah for both ZX81 and Spectrum, Micro Speech from Currah, Orator from Fuller, VoxBox from Datel Electronics, Chatterbox by William Stuart Systems, and S-Pack by DCP.

I contacted all of the above companies and those units that were supplied for review I sent to different reviewers for their independent comments.

For your own assessment, here's what they thought...

A neat black plastic box which fits snugly onto the back of the Spectrum and is identical in size and shape regardless whether it contains the Orator on its own or the complete Master Unit. It has a through bus allowing all the ZX peripherals to be attached, a cassette interface enhancing the signal and allowing both leads to be left in place when loading and saving, and a joystick interface.

The unit is supplied with a demonstration tape and what are alleged to be complete instructions. These are too verbose, complicated and ill-defined for the novice to understand easily. With its amplifier and larger speaer it gives the Spectrum a controllable audio output, although this requires a plastic screwdriver as no Knob is supplied - a metal one turns it into a radio! It can also be connected to your Hi-fi system.

The speech bit of it uses the G1-SP025A speech processor with a single chip N-channel MOS/LSI circuit that will synthesize speech via a stored program.

Speech is created by an allophone system which is accessed by utilising port 159. In practical terms each allophone is operated by sending numbers with OUT 159,n - where n is the appropriate numeric value assigned to that sound.

The unit is complex to program, step by step simple instructions would have helped, and it has the annoying habit of being incompatible with some commercial programs causing a crash or system reset or irritating unstoppable hiss.

The quality of sound can be very good, clear and distinct, but I'm afraid the programming takes up too much time for me to make frequent use of it.

The Sweet Talker unit (11 x 7.5 x 3cm) plugs easily and firmly into the user port at the rear of the computer and has its own rear card edge connector for a Ram Pack necessary. A demonstration program cassette is supplied and not only gives a good idea of the range of speech sounds possible but also explains how to program the unit and even gives a few ideas for turning games and original programs into 'talkies'. It takes less than a minute to load.

One way to understand how spoken words are formed is to break up the sounds that combine to make a word into allophones, and this unit uses the allophone system to produce words. For instance, the word Cheetah can be broken down to four speech sounds or allophones: /CH//IY//TT//AA/. Using the allophone system, each allophone, pause or gap in speech (as at the end of words and sentences) is turned into a number and then given a code. The code is then incorporated within a very simple program using OUT instruction to 7.

An instruction booklet is supplied with the unit and lists the allophones with everyday words which contain examples of the relevant allophones. Find the allophones required by checking through the word lists, then read across to the column listing the Data Codes and enter them in the first statement, followed by four more short lines and RUN. It talks...

There are several bonuses with this system. Firstly, although experimentation is necessary to achieve maximum clarity, it's great fun and often amusing. The booklet is easy to use. Secondly, it teaches the user a lot about the structure of words, vowels, consonants, word endings and so on. Not unlike the user of phonics in the teaching of reading. In this last respect this system could well find a place in classrooms. One word of warning to parents and teachers; the unit will say *anything* it's told to - and without so much as a blush!! But when will we able to program out micros to sing "Daisy, Daisy..."?

Unlike most speech units on the market today the DCP Speech unit does not create words nor does it use allophones.

The unit contains a pre-recorded chip which contains about 70 words which cannot be added to or changed.

The words have been created by first having been spoken and then changed into a digital form, they can be recalled by the user by simply using an OUT statement, i.e.: OUT 255,34. The advantage of this is that you get a clear reproduction of sound unlike the allophone system.

You can buy another three chips which contain another 70 words each, but will cost you another £12.95 each. These are just slotted into place inside the box. The DCP speech unit will fit the ZX81 and the Spectrum, and fits straight onto the output port.

You can also fit an external speaker if you wish, which can be adjusted by a volume control. The chip that comes with the unit contains the alphabet plus numbers from 1 to 20, and also the words hundred, thousand, million. So in fact you can construct any number you wish.

The other three chips contain words like control, equal, flow, gallon, high, inches, kilo, limit, lower, stop, emergency, pico, kilo and so on. This gives a total of about 280 words. Using the unit is very simple and it does add a new dimension to your programs, athough you can bet the one word you want won't be available.

The unit itself costs £29.95 and if you were to buy the unit and all the chips it will cost you nearly £70. Need I say more...?

This unit is one of the most easily obtainable, as many shops including Woolworths sell it. Packed in an eye-catching box, the unit itself is neat and lies flat on its back when connected to the user port, unlike most peripherals which stand upright. It is sold complete with a well written manual and demo tape which also contains an adventure game featuring speech.

A useful feature is that the sound, both speech and the normal Spectrum sound, are output through the TV speaker. A drawback is that the unit is dead ended and you cannot add anything on after it. I also found that the leads which are permanently attached to it are a little short if you want to add it on after other units. The unit uses an allophone system, but instead of the usual technical means of using the OUT instructions followed by the respective number, Micro Speech reserves S$ for its own use and a ROM in the unit decodes the required word.

Although the system is more related to English words, to get the correct pronunciation you have to break the word into its component syllables and reconstruct it in Currah language. "Hello," for instance, is spoken by using LET S$="HE(LL) (OO)" And of course there are ways to add information and inflexion to the speech, all of which make it a reasonably complex bit of equipment to use.

Currah have spent a great deal of time and effort persuading many software houses to adopt their unit for producing speech in their games, and this is obviously an incentive when purchasing such a unit. Most games also tend to use the Kempston joystick interface as a standard, and the two units cannot be operated at the same time, at least until now!

Currah once more show that they are aware of the market and its shortcomings and have produced an extension to the Spectrum's port which will allow two (or more) units to be used simultaneously. This can be used with other items such as printer interfaces of course.

After all this, how does it perform?

To be honest, I was disappointed at first. It didn't seem to live up to its promotion. The demo tape seemed indistinct and words were not pronounced clearly at all, a bit of very careful adjustment of the TV and the unit improved this slightly.

However after using it for a few days I became used to the electronic voice and began producing scathing comments to go with my games. The software companies who use it seem to get a better quality of speech and so far those who did use it used it sensibly and sparingly.

I must now admit that I find it an interesting and effective extension to my Spectrum and will make much use of it. Recommended to anyone who has an interest in going one step further.

Chatterbox II is quite a large unit - 10 x 15 x 6cm - and is housed in a black plastic case whch has two DIN sockets and a mini jack socket fitted on the top.

The two DIN sockets allow you to connect the WSS Speech recognition unit (£49.00 + VAT) to one, and either the output from the WSS Music Synthesiser and sixteen line control port (£19.50 Kit, £25.00 built + VAT) or output to an amplifier to the other. The mini jack is an input from the Spectrum's ear socket to amplify its beeps.

Compatible with *both* ZX81 and Spectrum, it is attached by a cable and plug similar to that on the ZX printer, this means that though the port is OK for ZX81 peripherals, very few Spectrum units will fit on the back.

Speech is achieved by using phonemes or individual speech sounds, and WSS themselves admit that the quality is not as good as pre-programmed speech, but add, and I see no reason to disagree, that the flexibility is infinite.

A list of these phonemes is given in the five page booklet which is brief but reasonably clear providing you are an experienced user. To allow for connection to other machines, a lot of technical detail is given, but unfortunately not how the INPUT/OUTPUT socket is wired so I couldn't try it with my Hi-fi. To actually produce speech, you have to break the word into its component sounds, look up the appropriate numbers and output them to port 127 using OUT 127,no.

The quality of sound depends on the amount of time that you spend in getting the phonemes right; even at its best it still sounds electronic, but clear and recognisable.

This will give hours of pleasure if you like tinkering, also it has great potential educational value as word structure has to be fully understood to get good results!