A technique for building up multiple tracks of music without using a multi-track audio recorder. In virtual mixing, performed parts are recorded as MIDI data on a sequencer. After all of the parts are recorded, each of the MIDI tracks is assigned to a different synth, or to a different layer on a multitimbral synth. The mix is automated by inserting MIDI volume and pan continuous controller data into the tracks. When all the tracks are complete, the whole thing is played back at once, and all of the audio outputs of the synths are mixed directly to stereo (or whatever the final playback format is to be).
The 1990s were the heyday of virtual mixing, when the performance/price ratio for computer hardware became just good enough that desktop computers could handle the recording and playback of complex MIDI tracks, and performers could afford enough synth capability to play all of the tracks needed in a song at the same time. Starting around 2000, as the desktop computers continued to improve, they became capable of recording and playing back multitrack audio; this eliminated the main motivation (lack of audio recording capability) for virtual mixing.