The idea is interesting – you take a sample and slice it into six new samples over frequency ranges from very low to very high. You can then shuffle the slices drawn from different samples mixing and matching their amplitude and/or frequency graphs. You might for (simple) example join a gong sound in the low frequencies and a violin in the high ones to make a gong-olin. To a certain extent you may create a hybrid of two (or as many as six) sounds, so long as you are able to join up the durations and range of their frequency spectrums.
Now obviously some sounds are long (organ) and others are sharp (anvil) and not everything will merge successfully, leaving you with something simultaneous but not coherent. Novum offers granular synthesis to blend things over shared duration more smoothly. That works – but everything tends to a similar wash under the amplitude graphs, like a smoothie. Many Novum sounds are drones or at least happen in large churches. There’s styles of music that benefit from this blender – the dark ambient folk will be well pleased. But many of the things I tried just became a blur.
Of course like any tool it takes time to get the gist of it. Random selections won’t help much – you have to plan to e.g. take the amplitude graph from a piano and apply it to a tuba in the most recognisable frequency domain. Impulsive people need not apply – so try out the demo and see if you are one of them. I would like to see this idea in spectral synthesis, granular is just too ‘glowing’.