Stereo Micro ↔ Macro Time Collapser — User Guide

Electroacoustic composition tool that inverts temporal scales: micro-bursts stretched into sustained textures, slow passages compressed into transient gestures. Stereo decorrelation via parallel granular processing.

Author: Shai Cohen Version: 1.0 (2025) Technique: Time-Scale Inversion + Granular Stereo Application: Praat scripting language

Contents:

What this does Quick start Time-Scale Inversion Theory Preset Strategies Parameters & Controls Visualization & Analysis Applications

What this does

This script implements a Stereo Micro ↔ Macro Time Collapser — an electroacoustic composition tool that performs temporal inversion: very short events (micro-bursts) are stretched into sustained textures, while long, slow passages are compressed into brief transient gestures. The result is a radical transformation of temporal perception, creating new relationships between time scales.

Key Features:

7 Preset Strategies — Gentle Bloom to Granular Chaos, plus Custom
Intensity-Based Segmentation — Detects micro-bursts (peaks) and slow textures (stable regions)
Time-Stretch Micro-Bursts — Overlap-add resynthesis with stretch factors up to 25×
Time-Compress Slow Textures — Compression factors as low as 0.05× (20× speed-up)
Stereo Decorrelation — Per-channel jitter (±stereo_width%) and pitch shift (±1-2%)
Multiple Interleave Modes — Alternate, Probabilistic, Timeline order
Target Duration Control — Generate output of specified length (up to 120s cap)
Comprehensive Visualization — 7-panel display showing analysis, segmentation, and results

⏱️ What is Time-Scale Inversion?

Time-scale inversion is a compositional technique that exchanges the roles of micro and macro temporal structures:

Micro → Macro: Very short events (10-200 ms) are stretched by factors of 5-25×, becoming sustained textures that reveal their internal structure
Macro → Micro: Long, slowly evolving passages (0.5-2.0+ seconds) are compressed by factors of 0.05-0.4×, becoming brief transients or gestures

The result: A composition where the original temporal relationships are inverted — what was brief becomes extended, what was extended becomes momentary. This creates a perceptual paradox, allowing listeners to experience sound at two different time scales simultaneously.

Technical Implementation: (1) Intensity Analysis: Extract intensity contour with short window (8-25 ms). (2) Micro-Burst Detection: Identify regions exceeding burst threshold with duration between min/max burst limits. (3) Slow Texture Detection: Identify regions with low intensity variance and shallow slope, duration > slow_min_ms. (4) Stereo Processing: For each channel independently, apply time-stretch/compress with per-channel jitter (±width%) and subtle pitch shift. (5) Interleaving: Assemble transformed segments according to selected mode. (6) Concatenation: Cycle through segment pool with crossfades until target duration reached. (7) Visualization: 7-panel display showing waveforms, spectrograms, segment maps, and duration statistics.

Quick start

In Praat, select exactly one Sound object (mono or stereo).
Run script… → select Stereo_Micro_Macro_Time_Collapser.praat.
Choose Preset (2-7 for specific strategies, 1 for custom).
Set analysis parameters (window sizes, thresholds, min/max durations).
Set transformation parameters (stretch/compress factors, crossfade).
Set stereo width percentage and interleave mode.
Specify target output duration (with maximum cap).
Enable Draw_visualization for comprehensive analysis display.
Click OK — processor analyzes, transforms, and creates stereo output.

Quick tip: Start with Gentle Bloom preset on a short (<10s) recording with varied dynamics. Enable visualization to see how micro-bursts (orange regions) and slow textures (blue regions) are detected. Listen to the output — you'll hear bursts stretched into pads and slow passages compressed into clicks/flurries. For dramatic effect, try Extreme Inversion or Granular Chaos. The output appears as "Stereo_Collapser_Output" in the Objects window.

Important: ANALYSIS ACCURACY depends on window_size relative to material. Percussive music: smaller windows (8-12 ms). Ambient/drone: larger windows (20-25 ms). STRETCH FACTORS > 20× may produce noticeable artifacts (granular texture). COMPRESS FACTORS < 0.1× may cause pitch shift artifacts (formants shift upward). STEREO WIDTH > 30% can cause extreme channel decorrelation (phase issues in mono). TARGET DURATION must be reasonable relative to source — extreme stretching may require many segment cycles. MAX OUTPUT DURATION hard cap prevents runaway processing.

Time-Scale Inversion Theory

The Inversion Principle

Given a sound with segments of varying duration: Micro-bursts: duration d_micro ∈ [min_burst, max_burst] Slow textures: duration d_slow ∈ [min_slow, ∞) Inversion transforms: d'_micro = d_micro × stretch_factor (where stretch_factor > 1) d'_slow = d_slow × compress_factor (where compress_factor < 1) Result: Original shortest events become longest Original longest events become shortest Temporal relationships inverted Example: Original: burst=50ms, slow=2000ms Stretch=12×, Compress=0.15× Transformed: burst=600ms, slow=300ms Now burst is longer than slow — temporal roles reversed

Segmentation Logic

🔍 Micro-Burst Detection

Criteria: intensity > median + burst_threshold_dB

Duration range: micro_burst_min_ms to micro_burst_max_ms

Peak tracking: continues while intensity stays within 3 dB of threshold

Markers: Peak intensity stored for visualization

🔍 Slow Texture Detection

Criteria (both must be true):

Variance threshold: std/mean < slow_variance_threshold (lower = more stable)
Slope threshold: |ΔdB/Δt| < slow_slope_threshold_dB_per_s (lower = less change)

Minimum duration: slow_min_ms (ensures texture length)

Stereo Decorrelation Mechanism

For each channel independently: 1. Duration Jitter: stretch_factor' = stretch_factor × (1 + random(-width/200, +width/200)) compress_factor' = compress_factor × (1 + random(-width/200, +width/200)) 2. Pitch Shift (subtle): f_shift = random(0.98, 1.02) for bursts f_shift = random(0.99, 1.01) for slow textures Implementation: • Override sampling frequency: sr × f_shift • Resample back to original sr Effect: • Left and right channels develop slightly different timing and pitch • Creates spacious stereo image • Prevents identical L/R signals (avoids mono collapse) • Width% controls amount of decorrelation

Interleave Modes

Mode 1: Alternate burst1, slow1, burst2, slow2, burst3, slow3, ... • Balanced alternation • Ensures both material types are heard regularly Mode 2: Probabilistic For each segment position, choose 50% burst / 50% slow • Unpredictable sequence • May favor one type depending on random distribution Mode 3: Timeline Order Sort all segments by original start time, preserving original temporal relationships • Maintains original order of events • Inversion affects duration only, not sequence

Crossfade Concatenation

When concatenating transformed segments: if both segments longer than 2× crossfade: Concatenate with overlap: crossfade_s seconds • Cosine-shaped crossfade • Smooths transitions between segments else: Concatenate directly (no overlap) Crossfade length limited to 30% of shorter segment Prevents artifacts when segments are very short

Preset Strategies

Preset 2: Gentle Bloom (subtle expansion)

🌸 Gentle Time Expansion

Window: 12 ms | Burst range: 20-150 ms | Threshold: +6 dB

Slow min: 800 ms | Variance: 0.12 | Slope: 5 dB/s

Stretch: 8× | Compress: 0.25× | Crossfade: 20 ms

Stereo: 20% | Interleave: Alternate

Character: Subtle transformation — bursts become gentle swells, slow passages become shorter gestures

Use on: Acoustic instruments, speech, any material needing subtle temporal manipulation

Preset 3: Extreme Inversion (dramatic scale swap)

⚡ Dramatic Role Reversal

Window: 10 ms | Burst range: 15-100 ms | Threshold: +10 dB

Slow min: 500 ms | Variance: 0.20 | Slope: 8 dB/s

Stretch: 20× | Compress: 0.08× | Crossfade: 8 ms

Stereo: 20% | Interleave: Probabilistic

Character: Extreme — 15 ms bursts become 300 ms textures; 500 ms textures become 40 ms transients

Use on: Percussion, found sound, experimental material

Preset 4: Micro Detail Focus (emphasize bursts)

🔍 Burst Detail Revealed

Window: 8 ms | Burst range: 15-80 ms | Threshold: +12 dB

Slow min: 1000 ms | Variance: 0.10 | Slope: 4 dB/s

Stretch: 15× | Compress: 0.3× | Crossfade: 5 ms

Stereo: 20% | Interleave: Alternate

Character: Micro-bursts stretched to reveal internal structure; slow textures moderately compressed

Use on: Percussion, transient-rich material, speech plosives

Preset 5: Macro Drone (emphasize slow textures)

🌫️ Slow Textures Preserved

Window: 20 ms | Burst range: 40-200 ms | Threshold: +5 dB

Slow min: 400 ms | Variance: 0.25 | Slope: 10 dB/s

Stretch: 6× | Compress: 0.05× | Crossfade: 25 ms

Stereo: 20% | Interleave: Timeline order

Character: Slow textures compressed to near-transients (20× speed-up); bursts moderately stretched

Use on: Ambient, drone, sustained material

Preset 6: Granular Chaos (high fragmentation)

🌀 Extreme Fragmentation

Window: 8 ms | Burst range: 10-60 ms | Threshold: +15 dB

Slow min: 300 ms | Variance: 0.30 | Slope: 15 dB/s

Stretch: 25× | Compress: 0.05× | Crossfade: 3 ms

Stereo: 20% | Interleave: Probabilistic

Character: Maximum transformation — bursts become long textures (250 ms from 10 ms), slow passages become clicks (15 ms from 300 ms)

Use on: Experimental sound design, granular synthesis source

Preset 7: Smooth Morph (minimal disruption)

✨ Gentle Morph

Window: 25 ms | Burst range: 50-250 ms | Threshold: +4 dB

Slow min: 1000 ms | Variance: 0.08 | Slope: 3 dB/s

Stretch: 5× | Compress: 0.4× | Crossfade: 30 ms

Stereo: 20% | Interleave: Alternate

Character: Minimal transformation — bursts become moderate swells, slow textures slightly shortened

Use on: Any material needing subtle temporal manipulation

Parameters & Controls

Analysis Parameters

Parameter	Type	Default	Description
Micro_window_ms	positive	15	Analysis window for intensity (ms)
Micro_burst_min_ms	positive	25	Minimum duration for burst detection (ms)
Micro_burst_max_ms	positive	180	Maximum duration for burst detection (ms)
Burst_threshold_dB_above_median	positive	8	dB above median intensity to trigger burst
Slow_min_ms	positive	600	Minimum duration for slow texture (ms)
Slow_variance_threshold	real	0.15	Max (std/mean) for stable texture
Slow_slope_threshold_dB_per_s	positive	6	Max intensity slope (dB/s) for slow texture

Transformation Parameters

Parameter	Type	Default	Description
Stretch_factor	positive	12	Time-stretch multiplier for bursts (>1)
Compress_factor	positive	0.15	Time-compress multiplier for slow textures (<1)
Crossfade_ms	positive	15	Overlap duration between segments (ms)

Stereo Parameters

Parameter	Type	Default	Description
Stereo_width_percent	positive	20	Decorrelation amount (0-50% typical)

Assembly Parameters

Parameter	Type	Default	Description
Interleave_mode	option	Alternate	Segment ordering strategy

Output Parameters

Parameter	Type	Default	Description
Target_output_duration_s	positive	60	Desired output length (seconds)
Max_output_duration_s	positive	120	Hard cap on output length
Allow_fallback_if_insufficient_segments	boolean	1	Use statistical fallback if detection fails
Draw_visualization	boolean	1	Generate 7-panel analysis display
Play_output	boolean	1	Audition after processing

Visualization & Analysis

7-Panel Display

Stereo Micro ↔ Macro Collapser Visualization: Panel 1: TITLE • Script name, input filename, preset, target duration Panel 2: ORIGINAL WAVEFORM • Gray waveform • Duration, channels, sample rate Panel 3: INTENSITY CONTOUR WITH SEGMENT MAP • Background: blue = slow textures, orange = micro-bursts • Red dotted line = burst threshold • Black intensity contour • Orange dots = burst peaks • X-axis: time, Y-axis: intensity (dB) Panel 4: ORIGINAL SPECTROGRAM • 0-5000 Hz spectrogram of input Panel 5: RESULT WAVEFORMS (STEREO) • Left channel (blue) and Right channel (orange) side by side • Shows actual output duration Panel 6: RESULT SPECTROGRAM (L channel) • 0-5000 Hz spectrogram of output Panel 7: SEGMENT DURATION DISTRIBUTION • Bar chart showing average durations: - Original bursts (orange) - Original slow textures (blue) - Transformed bursts (stretched, red) - Transformed slow textures (compressed, dark blue) Panel 8: PARAMETERS & STATISTICS (lower right) • Number of bursts/slow textures detected • Average durations • Stretch/compress factors, stereo width, crossfade • Interleave mode, target/actual duration Panel 9: LEGEND • Color key for all visualization elements

Reading the Segment Map

What the colors mean:

Orange regions: Detected micro-bursts (short, high-intensity events)
Blue regions: Detected slow textures (stable, slowly varying passages)
Red dotted line: Burst threshold = median intensity + threshold_dB
Orange dots: Peak intensity within each burst
Black line: Raw intensity contour (dB)

Pattern interpretation:

Many orange regions = transient-rich material
Large blue regions = sustained, drone-like passages
Mixed regions = varied dynamics
Gaps = below-threshold or too short for detection

Reading the Duration Bar Chart

Bars represent:

Burst (orange): Average original burst duration
Slow (light blue): Average original slow texture duration
B×Str (red): Average transformed burst duration (stretched)
S×Cmp (dark blue): Average transformed slow duration (compressed)

What to look for:

Compare heights: Inversion should make B×Str taller than original, S×Cmp shorter
Extreme inversion = B×Str much taller than original, S×Cmp much shorter
Balanced = both transformed durations similar

Applications

Electroacoustic Composition

Use case: Creating new relationships between time scales in acousmatic music

Technique: Extreme Inversion or Granular Chaos on diverse source material

Workflow:

Collect source recordings with varied dynamics (percussion, speech, environmental sounds)
Apply different presets to different sources
Layer multiple outputs for complex textures
Use as source material for further processing

Sound Design for Media

Use case: Creating unusual transitions, risers, impacts, and textures

Technique: Micro Detail Focus or Gentle Bloom on specific sounds

Applications:

Risers/builds: Stretch percussion hits (Extreme Inversion)
Impacts: Compress ambient drones into brief crashes (Macro Drone)
Transitions: Mix stretched and compressed material (Smooth Morph)
Background textures: Stretch environmental sounds (Gentle Bloom)

Experimental Music

Use case: Generating unpredictable, evolving material for improvisation

Technique: Granular Chaos with probabilistic interleave

Advantages:

Unpredictable segment order creates aleatoric structure
Stereo decorrelation adds spatial complexity
Target duration control generates fixed-length pieces

Educational Demonstration

Use case: Teaching time perception and granular synthesis concepts

Technique: Enable visualization, compare presets on simple sounds

Learning outcomes:

Understand relationship between duration and perceived time scale
See how intensity analysis segments sound
Hear effects of time-stretching and compression
Explore stereo decorrelation principles

Practical Workflow Examples

🎵 From Percussion to Pad

Goal: Transform drum loop into evolving texture

Settings:

Preset: Micro Detail Focus
Source: 10-second drum loop with varied hits
Stretch: 15× (bursts become 150-1200 ms swells)
Compress: 0.3× (silences/gaps become short)
Target: 60 seconds

Result: Drum hits stretched into pitched swells, creating continuous texture from rhythmic source

🎬 Film Transition: Whoosh to Impact

Goal: Create dramatic transition from sustained whoosh to impact

Settings:

Preset: Extreme Inversion
Source: 5-second whoosh + impact recording
Interleave: Timeline order (preserves original sequence)
Target: 30 seconds

Result: Whoosh stretched to 20 seconds (sustained), impact compressed to 40 ms (brief spike) — creates dramatic build and sudden resolution

🎚️ Ambient Generator from Speech

Goal: Create evolving ambient texture from spoken word

Settings:

Preset: Gentle Bloom
Source: 30-second spoken phrase
Stereo: 30% (wider decorrelation)
Target: 120 seconds

Result: Plosives become gentle swells, vowels become compressed gestures — speech becomes abstract, evolving texture with spatial movement

Troubleshooting Common Issues

Problem: No segments detected (numBursts=0, numSlowRegions=0)
Cause: Thresholds too strict or material unsuitable
Solution: Enable fallback mode, reduce burst threshold (4-6 dB), increase variance threshold (0.2-0.3), reduce slope threshold (3-4 dB/s)

Problem: Output much shorter than target
Cause: Too few segments, stretch/compress factors too extreme
Solution: Increase target duration (segments cycle), reduce stretch/compress extremes, enable fallback for more segments

Problem: Output sounds granular/artifact-heavy
Cause: Stretch factor too high (>20×), crossfade too short
Solution: Reduce stretch factor, increase crossfade (20-30 ms), use Smooth Morph preset

Problem: Stereo image collapses in mono
Cause: Extreme width (>40%) causing phase cancellation
Solution: Reduce stereo_width_percent (10-20%), or accept that extreme decorrelation may not sum perfectly

Problem: Processing extremely slow
Cause: Many segments, each requiring multiple resynthesis operations
Solution: Reduce target duration, reduce max segments by adjusting thresholds, use shorter source material

Advanced Techniques

Custom analysis strategies:

Very short bursts (glitches): micro_burst_min_ms = 5, micro_burst_max_ms = 30
Long textures (drone): slow_min_ms = 2000, slow_variance_threshold = 0.05
Aggressive detection: burst_threshold = 3 dB, variance = 0.3, slope = 20 dB/s
Conservative detection: burst_threshold = 12 dB, variance = 0.08, slope = 3 dB/s

Extreme parameter exploration:

Stretch only: compress_factor = 1.0 (no compression), interleave = Timeline
Compress only: stretch_factor = 1.0 (no stretching), interleave = Timeline
Mono output: stereo_width_percent = 0 (identical channels)
Max stereo: stereo_width_percent = 50, pitch shift ranges expanded in script
No crossfade: crossfade_ms = 1 (minimum, abrupt transitions)