Something else to consider: DAW audio is handled as 32bit floating point internally (1bit sign, 8bits exponent, 23bits fractional). Each magnitude (..., 10^-2, 10^-1, 10⁰, 10¹, 10², ...) has an equal precision of 24bits signed.

Examples:

• [0.1 to 1.0) has 23bits of precision
• [0.01 to 0.1) has 23bits of precision
• [-0.1 to -1.0) has 23 bits of precision
• Etc.

So each smaller magnitude distributes the precision across a smaller range of values. In the reverse direction, each larger magnitude distributes the precision across a larger range of values.

Examples:

• [1.0 to 10.0) has 23bits of precision
• [10.0 to 100.0) has 23bits of precision
• [-1.0 to -10.0) has 23bits of precision
• Etc.

Unfortunately, this means that when your audio exceeds 0db, each magnitude of volume increase spreads the precision across a larger number of values.

Examples:

• [0db to 1db) has 24bits of precision
• [1db to 10db) has 24bits of precision
• [10db to 100db) has 24 bits of precision
• Etc.

At the 10db to 100db range, the result should be an audibly decayed bit depth.

So essentially, when your volume exceeds 0db, any samples outside of the -1.0 to 1.0 range have a decreased precision. At least that's how I would expect it to work with my rudimentary understanding of IEEE single-precision floating point.

I am going to give it a test when I get home tonight:
Synth -> 100db Gain -> -100db Gain -> Master

So the entire song was run through a pitch-shifting algorithm (probably Elastique) by whoever uploaded it to get around copyright claims (Non pitch-shifted version).

So just pitch shift a regular piano sample up by a half step I guess?

That Korg is Virtual Analogue, i.e. a digital simulation of analog, so it may not be what you are looking for (assuming that you are specifically looking for something analog).

If it's not already on the to-do list, a waveshaper snapin with both symmetric and asymmetric custom draw-able curves would be fantastic for both Multipass and Snap Heap usage.

I've been reading up on wdl-ol, and hopefully I'll get motivated to put this idea together:

Keep a rolling FFT of the signal (num of bands = sample rate), with the lowest frequency band determined by chosen plugin delay amount (e.g. 60hz = 16ms), and then independently process every frequency from the FFT with:

• Amplitude Scale
• Phase Shift
• Compression with Attack/Release

Then sum the frequencies back together.

1-2 Samples -> Volume Shaper 4 -> Maximus -> Multipass + Disperser -> Reverb -> Maximus

If I can't pick a good single sample, I'll find a sample with a good transient/attack, another with a good tail, and ASD envelope both of them appropriately to avoid phase cancellation issues.

It has a very heavy feel, yet sounds surprisingly clean. Good job!

I completely forgot about this plugin until yesterday:
SlickHDR

It's basically the time-domain version of a multiband compressor.
There are three parallel compressors:

• Fast response one for transients
• Slow response one for overall tone
• Medium response one to smooth the differences

The result is surprisingly impressive.