TL;DR: an option for the user to specify time, timbre, and pitch qualities of samples could help point Genopatch toward consistently offering better solutions.

1. Ability to specify one-shot vs. sustained amplitude

Loading a clearly "one-shot" sample in Genopatch does not guarantee that it makes a one-shot sound. Often, when I give it a drum sample or a short foley recording, its solutions are all looped textures. The sound may seem to decay completely to niente, but if I hold the note long enough I'm surprised by the sound's recurrence.

2. Ability to specify one-shot vs. looping timbre / pitch motion

Suppose I load a sustained sound with a harmonically rich transient, like a pan pipe or a trumpet. In these cases, Genopatch usually offers sustained amplitude solutions (good), but the solutions often inexplicably veer up or down in pitch (bad mad envelope settings?) or loop through timbral changes rather than acting as one-shots. So, it would be nice to be able to specify whether such timbral changes have loop or one-shot behavior. Re; pitch, I've also noticed that Genopatch frequently uses the LFO to a sound that has a mild, one-shot fluctuation in pitch. I find it odd that Genopatch tries to solve with the LFO instead of mod envelope > oscillator pitch. (Is this due to the coupled nature of the envelopes? I imagine if the "ideal" amp envelope is too different in shape or length from the "ideal" mod envelope, Genopatch would resort to using the LFO.

3. Ability to specify sample pitch and solution pitch (or even just octave)

Genopatch has trouble with harmonically rich sounds in general. And even with Correct Tuning activated, I find that its solutions are often offset an octave from the correct sounding pitch; for example, playing Middle C sounds the pitch of the C below it. It's a hassle to get those patches to sound at the correct octave without just moving the pitch slider on the main page, which I prefer to keep at 0. Between the 2 oscillator pitches and the various FM controls, it's like moving a wheelbarrow full of frogs. Maybe there could be a script to alter every parameter related to pitch so that a patch can sound at a different pitch. Or maybe correct octave detection just requires further training of the neural net.

4. Ability to specify intensity of reverb

Genopatch seems to gravitate toward using reverb to solve for resonant bodies (drums and percussion in particular), which I think is great. But Genopatch also favors using reverb when solving very dry samples, such as a recording of a raw analogue oscillator, or an acoustic sound that's been scrubbed clean of reverberation. This seems odd.