XY Reverse Slicer

XY Reverse Slicer records what goes into it and plays each completed slice in reverse, crossfading one slice into the next. Short slices turn a vocal into a stutter. Long ones turn a chord into a swell that arrives instead of decaying. Between the two is a wide range of rhythmic chopping that follows whatever you feed it, because the slices come from the material rather than from a sequencer.

Two axes and a mix:

• X — slice length, 20 ms to 2 seconds, skewed so the middle of the pad is 250 ms.
• Y — texture, the crossfade between consecutive slices: from 1.5 ms, a hard cut, up to half the slice, which laps every slice over the next.
• Mix, equal power, so dry and wet sum to the same loudness as either alone.

Slices land on the grid

While the transport is playing, slice boundaries follow the song timeline rather than the moment you pressed play. Two things follow from that:

• A bounce matches what you heard. Rendering a section gives the same slicing as playing it, in any host.
• Looping and locating keep their place. Jumping around the arrangement re-anchors to the timeline instead of restarting the pattern wherever the playhead landed.

With the transport stopped it free-runs, so live input still slices.

Where it fits

On a vocal, for the stutter that turns a line into a hook. On a melody or chord loop, long, for reverse swells that lead into the next bar. On drums, short, for a scramble that still follows the groove. On a send with Mix low, as an effect that comments on the part rather than replacing it.

Tested before it ships

pluginval at strictness 10 passes on Windows, macOS and Linux, and auval passes for the Audio Unit. Mix at 0 % is the input bit for bit, and so is bypass. The crossfade holds level: the slice windows sum to unit power within 0.0000002, so where consecutive slices are unrelated the fully wet level holds within about 0.01 dB, at every sample rate tested from 22.05 kHz to 384 kHz. Where they are correlated the two windows do not sum to unity — that is the arithmetic of a cos/sin crossfade, not a fault — and at the extreme, with every sample inside a crossfade, the level can move by up to +2.1 or −4.4 dB. On a musical part at the shortest slices with the texture at maximum, the most measured is about half a decibel. Sweeping slice length and texture while it plays produces no step larger than the ordinary crossfade. Output is bit-identical from 1 to 5000 samples per block, including irregular block sizes. The reported tail is 4 s, the longest the effect can produce, so hosts stop trimming the reversed tail when the slice is long. There is no latency and no allocation on the audio thread.