I mainly develop audio and music software for Linux. I am a maintainer of, and long-time contributor to, the DSSI project:

I am the principal author and maintainer of the following projects:

Some demonstration MP3s comparing my hexter DSSI plugin and a Yamaha TX7

Here are some quick examples I threw together of how close (or not) hexter comes to a real DX7. In each one, hexter is panned hard right, and my TX7 (a DX7 in module form) is panned hard left. These were done using hexter 0.5.7; all but the last example will sound much the same with the current (1.0.1) version.

chopin.mp3

This is some Chopin with random patch changes thrown in. Both the TX7 and hexter are running with 16 voice polyphony. You can hear that both hexter and the TX7 click a bit when stealing voices. Also, hexter's quite a bit more graceful when receiving a patch change via sysex -- the TX7 twangs here transitioning into the "acoustic piano" patch.

clav_and_bass.mp3

This is actually two TX7s in the left channel, and two instances of hexter in the right channel.

e_piano.mp3

Here you can hear that while hexter is pretty close to the TX7, it doesn't quite get the modulation right, resulting in a slightly brighter or more midrangy sound.

mallet.mp3

organs.mp3

A couple more examples.

effects.mp3

This shows some of the patches that hexter 0.5.7 had the most trouble with (thanks to Frank Neumann for collecting these). Many effects patches depend on amplitude or pitch modulation, LFO, and unusual envelope generator timings. As of version 0.5.7, hexter lacked the DX7's AM and PM and LFO, and the odd envelope timings have so far resisted all efforts at prediction. Version 0.6.0 added many of these missing features, so hexter's emulation of these effects patches is much more accurate now.

All of the examples were generated from the same MIDI stream (both sys-ex patch dumps and notes). The TX7 was connected directly to my Delta 44, and both it and hexter were recorded simultaneously at 44.1kHz using ecasound and JACK. Audacity was used to trim and normalize the recordings (no effects other than fade-in and fade-out were used), and lame with '-h -b192' was used to encode the MP3s.