Cueing
When memory is asked, and with what
Cue modes
- Segment (default)
- At each segment boundary, the segment just played cues memory, and the echo plays during the next segment. This is the delay-like mode.
- Progressive
- From the Progressive Start slot (of 16 per segment) on, memory is re-cued at every slot with the part of the segment heard so far, compared with the same stretch of each stored trace. Traces that began like this one play along in step with you, and the echo sharpens as the segment unfolds. Before the start slot, the previous segment’s echo plays.
- Rolling
- No grid at all. Every Rolling Interval, the last Rolling Window of input is searched for at every position inside every stored trace (20 ms steps, refined to about 1 ms), and memory plays what followed the best matches, aligned with now. The search is spread across audio blocks with a fixed amount of work per block, so it never spikes the CPU; the status bar shows how far behind the live cue is. Traces must be longer than the window.
Cue Smoothing crossfades whenever a live cue changes the echo. Prediction plays memory ahead of the current position, so you hear what came next last time.
Cue source
| Source | Memory is cued by |
|---|---|
| Input | What you play. |
| Sidechain | Another track. In Live, set the plug-in’s Sidechain input to that track. Memory still records the main input. |
| Random | A random address each segment: memory dreams. |
| Frozen | The last live cue, held. Combine with Freeze Memory for a stable loop of recollection. |
| Echo Chain | The previous echo: each echo cues the next. See Sequences. |
Shaping the match
- Address chooses what is compared: the spectrum, the notes, the sound or the attacks. See Addresses.
- Feature Focus compares the Spectrum address Full, Rhythm only (loudness over time) or Timbre only (the average spectrum). With Rhythm, a new style with the same pulse recalls the old one.
- Recency favours recently stored traces.
- Self Match decides whether the segment just played may answer itself.
- Cue Gate stops quiet segments from cueing memory at all, so the noise floor doesn’t trigger echoes.
- Sequence Context adds the previous segment to the match; see Sequences.
- Activation Power, Similarity, Negative Activations and Echo Normalization are covered in How MINERVA works.