In 1985, Fairlight Instruments introduced the Voicetracker VT-5, a device that allows you to operate MIDI and control voltage synthesizers with your voice or any monophonic instrument, rather than with a MIDI keyboard. It was priced at US$2495 when it first came out.

The Voicetracker's 8MHz 68008 CPU analyzes an input signal to determine the pitch and harmonic content of the sound (such as dynamics and breathiness). It takes about 1ms plus several cycles of the incoming sound to determine the pitch. As you might guess, the pitch information is output as MIDI note values and analog control voltages, while the harmonic content analysis can be used to control a synthesizer's filters (among other things).

The Voicetracker can be used to "double" parts, either live or from tape. The original sound can be either enhanced with the doubling sound, or replaced completely by it. The Voicetracker can follow the original pitch exactly, or quantize it to semitone steps.

Like other Fairlight products, the Voicetracker was continually updated, so not all Voicetrackers have the same features. Within a year after its introduction, Fairlight added the capability to perform "intelligent harmonization" -- up to seven note harmony in a designated scale. You can create six harmony setups and switch between them in real time. This capability allows one singer to provide the sound of an entire choir. This is particulary effective if the synthesizer is patched through a vocoder, which imposes the singer's vocal characteristics on the synthesized sound.

Either microphone- or line-level sound sources can be fed into the Voicetracker. These signals are fed right back out unaltered, so that they can be sent to a mixing console. Also, an audio source can be fed into the Voicetracker's voltage controlled amplifier.

The digital inputs and outputs include MIDI in, out, and through. The Voicetracker can send MIDI note messages, along with key velocity, pitch bend, and other controls. There are also patchable footswitch jacks and a serial port for connection to a computer.

The analog outputs (pitch, gate, envelope, brightness, and purity) run from 0 to +5V, with the pitch output at 1 volt per octave. The Voicetracker also contains a built-in synthesizer, with dual DCO and DCF, which is used primarily for practice.

Although the box itself is a fairly attractive package, the video screen interface is a bit primitive looking. But perhaps that is not surprising considering that it was designed to appear on a relatively low resolution video (TV) monitor.

The Voicetracker is essentially a single-purpose computer, but Fairlight did not feel that they needed to include a monitor. Perhaps if the Voicetracker were produced today, there would be a built-in LCD display on the unit. The screen resolution is only 320x200 pixels, so an LCD of that size would probably not add too much to the cost.

When a video monitor (NTSC or PAL format) is connected to the Voicetracker the pitch information is presented in a scrolling display. The frequencies can be shown in reference to a simulated on-screen piano keyboard or a numerical scale. The scrolling speed and direction can be altered, a visual metronome can be enabled, and you can zoom in until the screen shows just two semi-tones.

I'm not sure whether the Voicetracker was an original idea created by the Fairlight engineers or whether the technology was acquired and expanded upon, although the VT-5 designation of the first available model leaves me wondering what happened to Voicetrackers 1 through 4...

Michael Topic, Former Project Lead for the Fairlight Voicetracker, was in charge of the development of the Voictracker through most of 1986. He contacted me and answered a few questions and filled in some names:

The idea was brought to Fairlight by a guy called David Warrender, who I seem to recall was a Californian. He had a patent for pitch extraction by autocorrelation of zero crossings and had licensed the technology to Roland/Boss to make guitar tuners. His idea was that you could, with the new Motorola processors, make this thing work in real time, and that was the basis of the concept.

Kim Ryrie more or less laid out the front panel and Peter Vogel was instrumental in arriving at the technical spec. Andrew Tolmie did lots of the early work of productising David Warrender's design, and Andrew Henderson wrote the majority of the rest of the software (with Andrew Tolmie).

I managed the project and wrote some of the software too. It was all in assembler. Wayne Connolly (I always get the spelling wrong ... apologies to Wayne) was initially the test technician, but went on to write software for it too. These are all interesting guys with great non-work stories to tell. :)

The reason it was called a VT-5 was, I think, because of the number of simultaneous analogue outputs that the unit produced. The five was not a version number. You can count the sockets on the back of it to see if I recall correctly or not.

Looking back at the list of analog outputs (pitch, gate, envelope, brightness, and purity), seems to confirm Michael's recollection! He continues:

I always had a love/hate relationship with the VT-5. It was amazing. There is no doubt about it. Even IRCAM in Paris, who had developed their own version of a Voicetracker in-house, had abandoned it in favour of the tracking accuracy and speed of the VT-5. It was, and is, one of the fastest and most accurate pitch to MIDI devices ever built.

I even once hooked up six of them to a hexaphonic pickup on a guitar to get a guitar synth, on Fairlight's time and premises. :) The experiment was disappointing. This leads me to my frustrations with the darn thing.

It turns out that the perceptual model that underlies the whole concept of the Voicetracker is too simplistic. Advances in thinking about musical performance and instruments since it was built have lead to new algorithms to track polyphonic inputs and to pick up on a variety of nuances which are common techniques in the repertoire of the most advanced guitar players, for example hammer-ons, pull-offs, and split harmonics.

VT-5's understanding of its inputs was just too simplistic to do anything useful with those performance frills and simply ignored them, which in turn annoyed the hell out of the virtuosi.

Of course, the processors of the day didn't have the grunt to do the math back then, but today's DSPs are perhaps three or more orders of magnitude more powerful. Sometimes I feel I would like to revisit the concept of the Voicetracker, with modern real time hardware and advanced algorithms. But who would fund such an adventure?

Anyway, thanks for keeping knowledge about this device alive.

Recently, Michael wrote again and added this:

In one of those terrible moments where one realises that not all credit is being given where credit is due, I forgot to mention one of the contributors to the Voicetracker project that made a significant difference.

Tom Stewart was the guy that figured out what to do with the harmony mode. Let me back track here. I was into Eventide Harmonizers but really hated that they had a fixed shift ratio, because it was an unmusical thing. Some of the harmonies were just not in key. (Bizarre aside: A man I worked for at another company after leaving Fairlight, Digital Audio Research, is now the head of R&D at Eventide, in one of those ironic twists).

I had this idea that you could use the Voicetracker as a MIDI harmony generator. Since we already had the current pitch information, you could theoretically generate real time harmonies that were in key. So, it was obvious. Specify the key and all would be well.

Then I needed to know how to generate those musically interesting harmonies. In stepped Tom Stewart, musicologist, technologist and all round clever and good guy. Tom was the person responsible for sampling some of the definitive, signature, and most oft-copied Fairlight CMI voices (orch5, saraaahh, etc) using a Series One CMI. He knew exactly how to generate a cool harmony, given the current pitch and the key. Thus the harmony mode was invented. I started agitating to have the thing become real, rather than just be an idea on paper, so Tom wrote down the ratios we needed, Andy Henderson coded it up in about a week (to shut me up, I think) and Voila!

I wrote some internal white papers on how this new feature could be used to drive a harmoniser and vocoder, but nothing much happened with that. The product literature didn't talk about the harmony mode, since it wasn't planned - it just sort of happened after all the product conceptualisation and brochures were done. We briefly thought about changing the VT5 menu structure, which would have meant scrapping all the front panels and screen printing them again. The cost of doing that made it out of the question.

We could have patented the pitch-sensitive harmony mode, I think, but there was an in-house policy of not doing that at the time (Fairlight did begin seeking patents later, though). If we had, it's my opinion that Fairlight may still have been collecting royalties on it, because many companies now provide pitch sensitive harmonisers. I think we were the first, though. (Don't throw law suits at me - I'm only vaguely recollecting!)

We certainly demonstrated the concept at NAMM 86 in Chicago (using a trumpet, a VT5 and a DX7 synth programmed to sound like a guitar, driving a Tom Scholtz Rockman rack - which I still have!). The sound we got was something like those dual lead guitar harmonies that the band Boston made famous. It was strange that a trumpet player could make that sound, but then Clive Smith, the trumpet player, was also a decent guitar player influenced by Hendrix, so that helped.

I looked up the patent for the VT5. It is listed on the www.uspto.gov site. The patent number was 4,429,609 issued on Feb 7th, 1984. David Warrender was Californian, as I had thought. He lived in Sebastopol at the time.

Anyway, I thought you might be interested to know how the harmony mode happened and who was responsible for making it sound so great.

You thought right!