This is really strange idea. I can image that it will have use. but Your description of what is “negative” is strange. Logical would be use something like FFT based waveform generation kind of like vocoder with inverted behaviour of band synthesis, because sound has more than volume dimmension. - If You just want to have volume dimmension - You can use side chained compressor which will create Level based cross fader between Your audio and 2nd track with whatever you want to be inverted sound. But it competely miss frequency spectrum inversion - This would be maybe possible with extremely hard coding of something like Halion 6 wavetable editor, where You can draw waveform as spectrum.