so, the script looks f’ing insane but i have not dug in yet
i’m not exactly sure what is described…
is it: around loop endpoints, in xfade period, there are volume “bumps” with repeated passes with nonzero pre level?
that is explored in this old issue:
in a nutshell, there is no good solution to this that doesn’t require expensive analysis of the input signal (i think) - there are only compromises. the default compromise gives you small volume “bump” with continuous high-intensity input. (see graphs in issue.) there are some “hidden” parameters in softcut that change the balance of this compromise, which we can get into if you want
other possibility: does the script manipulate rec_offset at all? i’d recommend just leaving that alone. currently it is possible to end up with a situation where you record with one offset, and then change the offset, and the old region will never be hit. here is the issue https://github.com/monome/norns/issues/898
but i wasn’t really committed to “fixing” this, because its a weird edge case (the real purpose of the offset is just to keep read/write heads out of each other’s resampling windows) and a fix would have performance impact.