Listening examples

The example presets, rendered offline

Each example is 16 bars at 120 BPM of generated test material (see Offline tools), processed by the engine with one of the presets in presets/examples/. Most are echo only (dry off), so what you hear is only what memory returns. The style change input plays one style for 8 bars, then another, which shows how memory handles the new material.

These are also in the plug-in’s preset menu under Examples; settings shown with @ change at that bar during the render (the plug-in ignores those).

The input

Style change (most examples)

Drums (short-trace and drum examples)

C major scale (scale-walk examples)

01 tape delay

Capacity 1: MINERVA reduces to a plain 1-bar tape delay (echo only).

capacity = 1 dry_level_db = -60

02 memory power3

100-trace memory, default activation power 3 (echo only).

capacity = 100 power = 3 dry_level_db = -60

03 memory power9

Power 9: the echo narrows to the closest memories (echo only).

capacity = 100 power = 9 dry_level_db = -60

04 memory power1

Power 1: every memory contributes - a blurred schema (echo only).

capacity = 100 power = 1 dry_level_db = -60

05 no self match

Self-match off: the cue cannot retrieve itself, so memory answers with the most similar earlier bars (echo only).

capacity = 100 power = 9 self_match = off dry_level_db = -60

06 encoding failure

Lf = 0.7: stored addresses are mostly forgotten; retrieval gets hazier (echo only).

capacity = 100 power = 3 encoding_failure = 0.7 self_match = off dry_level_db = -60

07 feedback mix

A playable setting: dry + echo, feedback 0.6, 100 traces.

capacity = 100 power = 3 feedback = 0.6 echo_level_db = -4

08 short traces

125 ms free-running traces: a granular-ish memory of short slices (echo only).

sync_mode = Free trace_ms = 125 capacity = 200 power = 5 self_match = off dry_level_db = -60

09 ternary familiarity

Classic MINERVA ternary features + familiarity-gated loudness, self-match off: the echo is loud only when the cue resembles earlier bars (echo only).

capacity = 100 feature_mode = Ternary power = 3 self_match = off normalization = Familiarity dry_level_db = -60

10 freeze after style a

Freeze memory after style A. When style B arrives, memory can only answer with style-A bars (echo only).

capacity = 100 power = 9 self_match = off dry_level_db = -60 @8.25 freeze = on

11 store novel only

Novelty gate: a bar is stored only if memory has nothing very like it. Memory stays small, a catalogue of distinct bars (echo only).

capacity = 100 power = 5 novelty_mode = Store Novel novelty_threshold = 0.7 self_match = off dry_level_db = -60

12 clamp opening

The first two bars are clamped into a 4-trace memory, so they never leave. Later bars cycle through the other two slots (echo only).

capacity = 4 power = 9 self_match = off clamp_incoming = on dry_level_db = -60 @2 clamp_incoming = off

13 merge prototypes

3-trace memory that consolidates instead of forgetting: each new bar is averaged into the most similar trace, so prototypes form (echo only).

capacity = 3 full_policy = Merge Similar power = 3 dry_level_db = -60

14 worn tape

Forgetting over time: traces fade, lose features and dull with age, with tape dropouts at recording (echo only).

capacity = 100 power = 3 self_match = off decay_fade_db = 1.5 decay_forget = 0.05 wear_tone = 0.8 content_dropout = 0.15 dry_level_db = -60

15 echo feeds memory

Record Source = Echo: after the first bar, memory stores what it plays back rather than what you play, so it keeps retelling its first memory (echo only).

capacity = 8 power = 3 record_source = Echo dry_level_db = -60

16 progressive in step

Progressive cue: from the 2nd sixteenth of each bar, memory is re-cued with the bar so far. Earlier bars that began the same way play along in step.

cue_mode = Progressive capacity = 100 power = 9 dry_level_db = -60

17 progressive prediction

Progressive cue with prediction: memory plays an eighth note (250 ms) ahead of you, so you hear what came next last time (echo only).

cue_mode = Progressive lookahead_ms = 250 capacity = 100 power = 9 dry_level_db = -60

18 rolling continuation

Rolling cue: no bar grid. Every 100 ms the last 250 ms of input is searched for inside every stored bar, and memory continues from the best matches (echo only).

cue_mode = Rolling rolling_window_ms = 250 rolling_interval_ms = 100 capacity = 100 power = 9 dry_level_db = -60

19 rolling drums mosaic

Rolling cue on drums with a short 120 ms window, self-match off: a mosaic of earlier hits that follows the groove (echo only).

cue_mode = Rolling rolling_window_ms = 120 rolling_interval_ms = 60 cue_smoothing_ms = 15 self_match = off capacity = 100 power = 6 dry_level_db = -60

20 three heads

Mode selector 1+2+3: like the RE-201’s three playback heads. Head 1 answers the last bar, heads 2 and 3 replay what head 1 played 1 and 2 bars earlier, panned left and right (echo only).

mode_selector = 1+2+3 capacity = 100 power = 9 dry_level_db = -60

21 iterative heads

Iterative heads: head 2 is cued by head 1’s echo, head 3 by head 2’s. Each generation drifts toward what memory holds in general (echo only).

mode_selector = Iterative 1+2+3 head2_level_db = -3 head3_level_db = -6 capacity = 100 power = 3 self_match = off dry_level_db = -60

22 chorus voices

Five strongest memories as separate chorus voices: spread across the stereo field, detuned and slightly delayed (echo only).

playback = Voices voices = 5 voice_spread = 0.9 voice_detune = 12 voice_delay_ms = 15 capacity = 100 power = 2 dry_level_db = -60

23 sampling dreams

Sample playback: each bar, one memory is drawn at random (weighted by activation). From bar 12 the cue is random, so memory dreams on its own (echo only).

playback = Sample capacity = 100 power = 3 self_match = off dry_level_db = -60 @12 cue_source = Random

24 rhythm focus

Feature focus = Rhythm: bars are matched by their loudness over time only, ignoring timbre, so style-B bars recall style-A bars with a similar pulse (echo only).

feature_focus = Rhythm capacity = 100 power = 9 self_match = off dry_level_db = -60

25 familiarity tone

Familiarity steers tone: echoes of familiar material are bright, echoes of the unfamiliar are muffled (echo only).

echo_tone_hz = 5000 intensity_to_tone = 1 capacity = 100 power = 3 self_match = off dry_level_db = -60

26 varispeed repeat rate

Varispeed: 8 bars are stored as 1-bar traces, then memory is frozen and the trace length (repeat rate) drops to half a bar. Old traces now play at double speed, an octave up, like turning up a Space Echo’s repeat rate (echo only).

capacity = 100 power = 9 length_mismatch = Varispeed dry_level_db = -60 @8.25 freeze = on @8.25 trace_division = 1/2

27 worn space echo

The classic machine: a 1-trace memory (plain tape delay) with wow, flutter, tape drive, hiss, darkening repeats and a spring tail. Dry + echo.

capacity = 1 feedback = 0.65 echo_level_db = -3 wow = 0.5 flutter = 0.4 tape_drive = 0.5 hiss_db = -58 feedback_bass_db = -3 feedback_treble_db = -6 spring_level_db = -12 spring_decay = 2.5

28 spring only

RE-201 mode 12: spring reverb only, on the dry signal, no echo.

echo_level_db = -60 spring_level_db = -4 spring_on_dry = on spring_decay = 3

29 memory on tape

The memory echo with tape character: 100 traces, feedback through a darkening EQ, gentle wow and flutter, drive and a spring. Dry + echo.

capacity = 100 power = 3 feedback = 0.5 echo_level_db = -3 wow = 0.3 flutter = 0.25 tape_drive = 0.35 hiss_db = -64 feedback_treble_db = -5 spring_level_db = -14

30 spectral blend

Power 1 blends every memory, as in 04, but as spectra: no phasey comb filtering from unaligned bars. Compare with 04_memory_power1 (echo only).

blend_domain = Spectral spectral_voices = 16 capacity = 100 power = 1 dry_level_db = -60

31 stretch repeat rate

As 26, but Length Mismatch = Stretch: halving the repeat rate squeezes the old bars into half the time without changing their pitch (echo only).

capacity = 100 power = 9 length_mismatch = Stretch dry_level_db = -60 @8.25 freeze = on @8.25 trace_division = 1/2

32 spectral freeze drone

Spectral freeze: at bar 6 the echo’s spectrum is held as a drone for four bars, then released (echo only).

blend_domain = Spectral capacity = 100 power = 3 dry_level_db = -60 @6 spectral_freeze = on @10 spectral_freeze = off

33 grain cloud

Grain memory: 40 ms traces, up to 2000 of them. Each grain is answered by the 24 most similar earlier grains, excluding itself: a granular cloud (echo only).

sync_mode = Free trace_ms = 40 capacity = 2000 max_active = 24 power = 6 self_match = off dry_level_db = -60

34 anticipation

Sequence context, Predict Next: each bar is answered by what followed similar bars before. Once style A repeats, the echo anticipates the next bar (echo only).

sequence_context = on context_cue = Predict Next capacity = 100 power = 9 dry_level_db = -60

35 sequence match

Sequence context, Match Both: memory recalls the bar that matches both the bar just played and the one before it, not just similar-sounding bars (echo only).

sequence_context = on context_cue = Match Both capacity = 100 power = 9 self_match = off dry_level_db = -60

36 memory sequencer

Echo Chain, free-running: from bar 8 each step plays one remembered bar, drawn by how well its context matches the last, and that bar cues the next. Memory walks through style A on its own while style B plays in (echo only).

sequence_context = on context_cue = Predict Next capacity = 100 power = 9 dry_level_db = -60 @8 freeze = on @8 cue_source = Echo Chain @8 chain_input = 0 @8 chain_step = Sample @8 playback = Sample @8 habituation = 0.4

37 steered chain

Echo Chain steered by the input (Chain Input 0.5): memory follows its learned sequence, pulled toward what is played now (echo only).

sequence_context = on capacity = 100 power = 9 dry_level_db = -60 @8 freeze = on @8 cue_source = Echo Chain @8 chain_input = 0.5

38 scale walk spectrum

The scale input (C to C’ and back to C, a note per bar), then the echo chain walks memory on its own from bar 9 with Address = Spectrum: the spectrum barely tells neighbouring notes apart, so the walk drifts.

sequence_context = on context_cue = Predict Next address = Spectrum capacity = 100 power = 6 echo_level_db = -4 @9 freeze = on @9 cue_source = Echo Chain @9 chain_input = 0 @9 chain_step = Sample @9 playback = Sample @9 habituation = 0.4 @9 cue_noise = 0.15

39 scale walk pitch class

The scale input (C to C’ and back to C, a note per bar), then the echo chain walks memory on its own from bar 9 with Address = Pitch Class: each note’s successor is recalled, so the walk plays the scale (octave-blind: C’ may be taken for C).

sequence_context = on context_cue = Predict Next address = Pitch Class capacity = 100 power = 6 echo_level_db = -4 @9 freeze = on @9 cue_source = Echo Chain @9 chain_input = 0 @9 chain_step = Sample @9 playback = Sample @9 habituation = 0.4 @9 cue_noise = 0.15

40 scale walk pitch

The scale input (C to C’ and back to C, a note per bar), then the echo chain walks memory on its own from bar 9 with Address = Pitch: each note’s successor is recalled, so the walk plays the scale, C’ included.

sequence_context = on context_cue = Predict Next address = Pitch capacity = 100 power = 6 echo_level_db = -4 @9 freeze = on @9 cue_source = Echo Chain @9 chain_input = 0 @9 chain_step = Sample @9 playback = Sample @9 habituation = 0.4 @9 cue_noise = 0.15

41 same sound

Address = Timbre: bars are recalled by what they sound like, not which notes they play. Style A’s bright pad and style B’s darker stabs each recall their own kind from memory, whatever the chord.

address = Timbre capacity = 100 power = 9 self_match = off dry_level_db = -60

42 groove recall drums

Address = Rhythm, 1/4-note traces over the drum loop: each beat recalls beats with the same attack pattern (kick, snare or hats in the same places), whatever else is going on.

address = Rhythm trace_division = 1/4 capacity = 200 power = 9 self_match = off dry_level_db = -60