From Jazz Player, February/March 1996

Jazz Fusion

By Bob Johnson

Have you ever wished as a wind instrument player or guitarist that you could play a MIDI instrument other than a keyboard that was close in its fingering and response to your native instrument? Over the past decade or so since MIDI arrived, several solutions to this intriguing problem have appeared on the market. The first group fall into the category of alternate MIDI controllers, such as the Akai EWI (electronic woodwind instrument) and EVI (electronic valve instrument), the Yamaha WX7 and WX11 wind controllers and even various percussive controllers like the trap KAT MIDI drum kit complete with bass drum and highhat pedals. Guitarists have also seen the emergence of MIDI guitar pickups.

The second group of solutions are the different forms of pitch to MIDI converters, one of which we will examine in detail here in this article. Previously available, but no longer being manufactured were the IVL Pitchrider and the Roland VP-70. Both of these were dedicated pieces of hardware that accepted a monophonic audio signal, say from a microphone, and would output a corresponding MIDI stream that could be fed into any MIDI module. Now on the scene for Macintosh or Windows PC’s is a capable piece of software called "Autoscore" from Wildcat Canyon Software of Berkeley, California.

What it is

I worked with the PC Windows version, so I’m assuming most of this goes for Macs as well. Installation went very quickly for me without a hitch. After you have finished installing the program, you will notice an extra item in the top menu bar of the Program Manager. In fact this extra Autoscore item appears automatically in every MIDI sequencer program that you run from Windows. The deal is, Autoscore adds its menu to these "host" programs. Then when you feed an audio signal of a single non-chordal instrument into your Windows compatible sound card, Autoscore sends a corresponding MIDI stream to the host sequencer to record as input. Included with Autoscore is a basic level sequencer, "Recording Sessions" from Midisoft, that is totally compatible, so you can work there and export to other sequencers via Standard MIDI files (SMF’s).

Program options for the package are all chosen via dialog boxes accessible from the Autoscore menu. The most important dialog box has to do with instrument settings that Autoscore will use to perform its pitch to MIDI algorithm. This covers things like the range of the input instrument (voice, guitar, wind/brass, string or rhythm), the volume threshold and "average note length." Another important dialog covers the tracking options such as constraining the MIDI output to chromatic notes or else to a specified key, and toggling pitch bend and volume tracking on or off. Also, there is a dialog to control interface issues such as MIDI output channel, the MIDI driver to use, etc.

Autoscore is not intended as a live performance package (would you really bring your computer to a gig anyway?) and even in the studio, it is suggested in the manual that you turn the MIDI through switch in your sequencer off. Thus, you don’t immediately hear the MIDI equivalent of your efforts echoed as you play your own instrument.

Method of Testing

I tried to evaluate the program in a couple of discrete phases to see if I could nail down the full extent of its abilities. One of the abilities you would like to see in such a product is repeatability. This meant having an input for each test that was totally repeatable. To do this, I first selected a saxophone etude in 4/4 at a tempo of 120 beats per minute loaded with 16th notes. Instead of playing it myself and having variations from time to time, I had my synthesizer play the etude using its crude saxophone voice from within Recording Session and also ran the audio output from the synth into the input of my sound card. Then I simultaneously recorded the MIDI output from Autoscore in Recording Session (the host program). The options had been set to constrain Autoscore’s algorithm to chromatic notes with no pitch bending and no volume tracking. Lastly, I compared the original MIDI sax track to the one generated by Autoscore. The results were very good for this test. Exporting the test via SMF to my usual sequencer and examining the MIDI events of both tracks showed that the Autoscore track lagged behind the original by only six or seven clock ticks (at 240 ticks per beat) and that none of the notes were dropped. If you played both tracks simultaneously, there was no perceptible difference in the start times of all the notes. The only discrepancy seemed at the end of some phrases, where Autoscore was a little lax in cutting off the final note of the phrase. No big deal. What this tells us is that for a very regular almost mechanical style of playing with no pitch bending and with perfect intonation, Autoscore is about as close to 100% pitch to MIDI as you can get. To verify repeatability, I ran the test a number of times always with the same good results.

The next test involved taking recordings of my own playing on tenor and alto saxes. I have digital audio wave files of solo saxophone that I played through my sound card and then recorded on tape. You need to go to tape here because Autoscore ties up your sound card to do its own thing, hence, you can’t play the audio directly over the sound card into the program, but small matter. Feeding the tape recordings into Autoscore still gave decent results if you turned the pitch bend and volume tracking options off. Autoscore picked up say 85 to 90% of the proper notes, even glisses and rapid runs, but none of the breath control of course could be captured. With the pitch bend and volume tracking options turned on, Autoscore sometimes got lost during a phrase and would seem to spit out some rehash of a previous bar or two. Trying to adjust the note length and volume threshold options of Autoscore didn’t seem to help much either. Still, depending on what you intend to use the program for, it has merit.

The Verdict

It seems improbable that an after the fact pitch to MIDI processor could ever be as precise and expressive as a wind controller, which is essentially an elaborate system of switches and a breath controller hardwired directly into the MIDI universe. However, Autoscore performed extremely well under some conditions and has a couple of advantages that a wind controller doesn’t. For one you don’t have to develop chops on another instrument (the wind controller). Sure the wind controller is close to the other wind instruments in its concept, but it is different and requires even more precision in the fingering being less forgiving than a resonating tube of air, surrounded by brass or wood.

The second advantage could be even more significant in a fusion recording scenario. You can use Autoscore to layer the original acoustic sound contained within the digital recording of an instrument. To give you an example, a hiphop tune of mine originally had a tenor saxophone solo recorded in a wav file on my PC. I used Autoscore to create a MIDI copy of the solo and then played back the tune while doubling the acoustic solo on my synth. Even with 10 to 15% of the MIDI notes dropped, I could go back and edit most of the discrepancies out in my sequencer. Playing back both the saxophone and MIDI accordion tracks simultaneously gave an unusual, but interesting sound that definitely worked in the tune (hiphop meets zydeco). In addition, because the solo was doubled like this, none of the pitch to MIDI discrepancies really stood out. In fact, it introduced some randomness that made it sound more than one guy was playing the solo in unison. You could even fly the MIDI file into Band-in-a-Box to generate harmony parts to a solo if a tune so warranted it. Bottom line is, in the right hands this program could be dangerous.

back