Percussive Groove Pattern Creator — User Guide
Automatic drum pattern generation: detects bass drums, hi-hats, and snares from audio and creates algorithmic groove patterns with customizable styles.
What this does
This script implements automatic percussive groove creation — an intelligent approach to rhythm generation that analyzes audio files to detect percussive events (bass drums, hi-hats, snares) and reassembles them into musically coherent drum patterns. The process: (1) Audio Analysis: Detects sharp amplitude onsets and classifies them by spectral content. (2) Event Classification: Groups detected sounds into three categories: bass drums (low-frequency dominant), hi-hats (high-frequency dominant), and snares (mid-frequency dominant). (3) Pattern Generation: Arranges classified sounds according to selected beat patterns (standard 4/4, funk, breakbeat, etc.) with adjustable density and timing variations. (4) Dynamic Processing: Applies envelope shaping, velocity variation, and optional stereo imaging.
Key Features:
- Intelligent Detection — Automatically identifies bass drums, hi-hats, and snares from any audio
- 6 Beat Patterns — Standard 4/4, Syncopated Funk, Breakbeat, Half-time, Double-time, Sparse Minimal
- Customizable Density — Control pattern complexity from sparse to dense
- Dynamic Shaping — Adjustable attack/release envelopes and velocity variation
- Stereo/Mono Output — Create mono patterns or stereo variations
- Flexible Length — 1-bar, 2-bar, or 4-bar patterns
- Tempo Control — Set any BPM from 60-240
What are algorithmic groove patterns? Traditional drum programming: manual placement of samples on a grid. Algorithmic groove generation: automatic pattern creation based on detected sounds and musical rules. Advantages: (1) Time-saving: Automatically extracts usable sounds from any audio source. (2) Creative inspiration: Generates unexpected patterns from source material. (3) Consistency: Maintains musical coherence within selected style. (4) Adaptability: Works with any percussive audio material. (5) Variation: Creates unique patterns each time with controlled randomness. Use cases: Music production (quick drum tracks), sound design (rhythmic textures), sample library creation (extracting drum hits), experimental composition (algorithmic rhythms), education (rhythm pattern studies).
Technical Implementation: (1) Onset Detection: Analyze intensity contour, detect sharp amplitude peaks above threshold. (2) Spectral Classification: For each detected onset, analyze frequency band energies (low: 20-250Hz, mid: 250-4000Hz, high: 4000-18000Hz) to classify as bass drum (low dominant), hi-hat (high dominant), or snare (mid dominant). (3) Pattern Logic: Based on selected pattern type, probabilistically place sounds on sixteenth-note grid according to musical rules for that style. (4) Dynamic Processing: Apply customizable attack/release envelopes with shape control, random velocity variation (60-100%). (5) Assembly: Mix processed sounds into final pattern, with optional stereo variation for width. Key insight: The script doesn't require pre-classified samples — it intelligently extracts and categorizes percussive sounds from any audio material.
Quick start
- In Praat, select any Sound object containing percussive content.
- Run script… →
percussive_groove_creator.praat. - Set Pattern_length: 1 bar (simple), 2 bars (standard), or 4 bars (complex).
- Choose Beat_pattern: Start with "Standard 4/4" for familiar patterns.
- Set Tempo (BPM) to match your project (default 120).
- Adjust Onset_threshold to control detection sensitivity (default -20dB).
- Set Groove_density to control pattern complexity (0.1=sparse, 1.0=dense).
- Enable Create_stereo for wider, more interesting patterns.
- Click OK — pattern is generated, named "originalname_groove_Xbar".
Quick tip: Start with drum loop or breakbeat audio for best results. Use Onset_threshold = -25 to -15 dB to fine-tune detection (lower = more sensitive). Set Groove_density = 0.4-0.7 for natural-sounding patterns. Enable Create_stereo for more interesting results (creates slight variations between channels). Adjust Clip_max_length to control how long each sound plays (shorter = tighter, more defined). Use Attack_time = 0.002s and Release_time = 0.05s for punchy sounds. Pattern analysis shows detected counts of each sound type — aim for at least 3-5 of each type for best variation.
Important: SOURCE AUDIO QUALITY MATTERS — Clean, isolated percussive hits yield best classification. Complex mixes may produce misclassifications. Onset detection sensitivity critical — too high may miss sounds, too low may capture non-percussive transients. Not all audio will contain all three sound types — patterns use only detected types. Groove_density significantly affects pattern feel — high values create busy patterns, low values create sparse/groovy patterns. Tempo changes affect pattern duration — faster tempos = shorter patterns at same bar length. Stereo variation creates different patterns in left/right channels — not just panning. Random elements mean patterns vary each run even with same settings.
Percussive Event Detection
Onset Detection Algorithm
🔊 Intensity-Based Peak Detection
Process:
- Convert audio to intensity contour (70Hz analysis, 0 smoothing)
- Scan for local maxima in intensity
- Compare current value to previous and next frames
- Threshold at user-defined dB level (-20dB default)
- Enforce minimum silence between events (0.05s default)
Mathematical check: event detected if:
intensity[frame] > intensity[frame-1] AND
intensity[frame] > intensity[frame-2] AND
intensity[frame] > intensity[frame+1] AND
intensity[frame] > onset_threshold
Spectral Classification
Three-Band Energy Analysis
For each detected onset segment:
Extract segment: ±0.005s before, up to max_segment_duration after
Convert to spectrum
Calculate band energies:
LOW (20-250Hz): bass drum range
MID (250-4000Hz): snare/body range
HIGH (4000-18000Hz): hi-hat/cymbal range
Total energy = LOW + MID + HIGH
Calculate ratios:
low_ratio = LOW / total_energy
mid_ratio = MID / total_energy
high_ratio = HIGH / total_energy
Classification rules:
IF low_ratio > 0.55 → BASS DRUM
ELSE IF high_ratio > 0.35 → HI-HAT
ELSE IF mid_ratio > 0.4 → SNARE
ELSE → reject (non-percussive/ambiguous)
Why These Frequency Ranges?
Spectral characteristics of drum sounds:
- Bass Drums: Energy concentrated below 250Hz (fundamental and harmonics)
- Snares: Broad midrange emphasis (250-4000Hz for body and snare wires)
- Hi-hats/Cymbals: Bright, high-frequency content above 4000Hz
- Toms: Often fall between bass and snare ranges
Classification tips:
- Clean sources: Isolated drum hits classify most accurately
- Mixed audio: May require threshold adjustments
- Electronic vs acoustic: Electronic drums often have purer spectral profiles
- Room sound: Reverb can blur frequency boundaries
- If misclassification occurs: Try different source audio or adjust Max_segment_duration
Detection Parameters
| Parameter | Default | Effect | Typical Range |
|---|---|---|---|
| Onset_threshold | -20 dB | Detection sensitivity | -30 (sensitive) to -10 (strict) |
| Min_silence_between_events | 0.05 s | Prevents double-triggers | 0.02 to 0.2 s |
| Max_segment_duration | 0.15 s | Length analyzed for classification | 0.05 to 0.3 s |
Detection troubleshooting:
- Too few events detected: Lower onset_threshold, check audio has clear transients
- Too many false positives: Raise onset_threshold, increase min_silence_between_events
- Poor classification: Adjust Max_segment_duration to capture full attack portion
- Missing sound types: Source audio may not contain all drum types
Beat Pattern Algorithms
🎵 Pattern Generation Logic
Core principle: Each pattern type has predefined rules for which sounds to place on which sixteenth-note positions, with probability controlled by groove_density parameter.
Grid: Pattern divided into bars × beats × sixteenths
Placement: For each sixteenth position, algorithm determines if a sound should be placed and what type (bass, snare, hi-hat)
Pattern 1: Standard 4/4
BASS: Beat 1 (always), Beat 3 (probability = groove_density)
SNARE: Beats 2 and 4 (probability = groove_density + 0.2)
HI-HAT: On beats (1,3) and off-beats (probability = groove_density × 0.8)
Other sixteenths (probability = groove_density × 0.3)
Example at groove_density = 0.6:
Bass: 100% on 1, 60% on 3
Snare: 80% on 2 and 4
Hi-hat: 48% on beats, 18% on other sixteenths
Result: Traditional rock/pop pattern
Pattern 2: Syncopated Funk
BASS: Beat 1 (always)
Beat 2, sixteenth 4 (probability = groove_density) - anticipatory
Beat 3, sixteenth 3 (probability = groove_density) - syncopated
SNARE: Beats 2 and 4 (probability = groove_density + 0.2)
HI-HAT: Most positions (probability = groove_density × 0.9)
Character: Off-beat bass drum accents, constant hi-hat
Example: James Brown, Parliament-Funkadelic style
Pattern 3: Breakbeat
BASS: Beat 1 (always)
Beat 1, sixteenth 3 (probability = groove_density × 0.6) - double hit
Beat 3, sixteenth 1 or 4 (probability = groove_density)
SNARE: Beat 2 (always)
Beat 4, sixteenth 1 or 3 (probability = groove_density)
HI-HAT: Moderate density (probability = groove_density × 0.7)
Character: Complex, syncopated, inspired by classic breakbeats
Example: Amen break, Funky Drummer style patterns
Pattern 4: Half-time Feel
BASS: Every 2 bars on beat 1 (50% slower feel)
SNARE: Every 2 bars on beat 3 (probability = groove_density)
HI-HAT: On beats only (probability = groove_density × 0.6)
Temporal feel: Pattern spans twice as many bars
Character: Laid-back, hip-hop, dub
Example: 70-85 BPM hip-hop grooves
Pattern 5: Double-time Feel
BASS: On beats and off-beats (probability = groove_density)
SNARE: On off-beats (probability = groove_density × 0.8)
HI-HAT: Most positions (probability = groove_density)
Character: Fast, busy, energetic
Temporal feel: Pattern feels twice as fast as tempo suggests
Example: Jungle, drum & bass, fast punk
Pattern 6: Sparse Minimal
BASS: Beat 1 (always), Beat 3 (probability = 0.3)
SNARE: Beat 2 (probability = groove_density × 0.5)
HI-HAT: Every other beat (probability = groove_density × 0.4)
Character: Minimal, spacious, atmospheric
Density: Very low even at high groove_density settings
Example: Ambient, minimal techno, dub techno
Probability and Randomization
For each sixteenth position:
Generate random number: r = randomUniform(0, 1)
IF pattern rule says "place type X with probability p"
THEN IF r < p → place sound of type X
Example: Standard 4/4, groove_density = 0.7, beat 3 snare placement
Rule: snare on beat 3 with probability = groove_density + 0.2 = 0.9
Random number: r = 0.65
Since 0.65 < 0.9 → place snare
Result: Consistent pattern structure with natural variation
Parameters & Settings
Pattern Parameters
| Parameter | Type | Default | Description |
|---|---|---|---|
| Pattern_length | option | 2 bars | 1, 2, or 4 bars |
| Beat_pattern | option | Standard 4/4 | 6 pattern styles |
| Tempo (BPM) | real | 120 | Beats per minute |
| Groove_density | real | 0.6 | 0.1 (sparse) to 1.0 (dense) |
| Create_stereo | boolean | yes | Create stereo pattern |
Detection Parameters
| Parameter | Default | Description |
|---|---|---|
| Onset_threshold | -20 dB | Minimum intensity for detection |
| Min_silence_between_events | 0.05 s | Prevents double triggers |
| Max_segment_duration | 0.15 s | Maximum length analyzed |
Sound Processing Parameters
| Parameter | Default | Description |
|---|---|---|
| Clip_max_length | 0.12 s | Maximum sound duration |
| Attack_time | 0.002 s | Envelope attack time |
| Release_time | 0.05 s | Envelope release time |
| Shape_intensity | 1.2 | Envelope curve shape |
Parameter Interactions
Important relationships:
- Groove_density × Pattern_length: Longer patterns with high density can become overwhelming
- Tempo × Clip_max_length: Faster tempos may require shorter clips
- Attack_time × Sound material: Very short attack times (0.001s) work best with sharp transients
- Onset_threshold × Source loudness: Adjust relative to your audio's peak level
- Pattern_type × Groove_density: Some patterns (like Sparse Minimal) maintain sparseness even at high density
Workflow Examples
Quick Drum Track Creation
🎬 From Breakbeat to Custom Pattern
Goal: Create unique drum pattern from classic breakbeat
Source: "Amen Brother" or similar breakbeat sample
Settings:
- Pattern_length: 2 bars
- Beat_pattern: Breakbeat
- Tempo: 140 BPM
- Groove_density: 0.7
- Onset_threshold: -22 dB
- Create_stereo: yes
Result: New breakbeat-inspired pattern with similar feel but unique arrangement
Experimental Texture Creation
🎵 From Field Recording to Rhythm
Goal: Create rhythmic pattern from non-musical sounds
Source: Field recording of construction site, kitchen sounds, etc.
Settings:
- Pattern_length: 4 bars
- Beat_pattern: Sparse Minimal
- Tempo: 90 BPM
- Groove_density: 0.4
- Onset_threshold: -18 dB (more selective)
- Clip_max_length: 0.2 s (longer textures)
Result: Ambient rhythmic texture with found sounds
Drum Sound Library Organization
📚 Categorizing Random Drum Hits
Goal: Automatically sort folder of drum samples
Source: Folder of unlabeled drum hits
Workflow:
- Concatenate all samples into one long sound
- Run script with very low Groove_density (0.1)
- Script detects and classifies each hit
- Output shows counts of each type
- Manually verify and organize based on classification
Result: Preliminary categorization of unknown samples
Common Workflows
| Use Case | Recommended Settings | Tips |
|---|---|---|
| Traditional Rock | Pattern 1, Density 0.6-0.8, 2 bars | Use drum kit recordings |
| Hip-Hop | Pattern 4, Tempo 70-90, Density 0.5 | Add swing manually after |
| Techno | Pattern 1 or 5, Tempo 120-140, Density 0.8 | Use electronic drum sounds |
| Ambient | Pattern 6, Tempo 60-80, Density 0.3 | Long attack/release times |
| Jazz | Pattern 2, Density 0.5-0.7, Stereo yes | Use acoustic kit recordings |
Advanced Techniques
Layering Patterns
Creating complex rhythms:
- Generate multiple patterns from same source with different settings
- Mix patterns together in Praat or DAW
- Example: Layer sparse pattern (density 0.3) with dense pattern (density 0.8)
- Result: More complex, evolving rhythm
Tempo Automation
Creating tempo changes:
- Generate pattern at base tempo
- Use Praat's "Lengthen (overlap-add)" to change tempo without pitch
- Or generate multiple patterns at different tempos and crossfade
- For gradual tempo changes: Generate long pattern, then apply time-stretching
Hybrid Pattern Creation
Mixing pattern types:
- Generate pattern A (e.g., Standard 4/4)
- Generate pattern B (e.g., Breakbeat)
- Extract bass from A, snares from B, hi-hats from both
- Reassemble in Praat using matrix operations
- Result: Hybrid pattern with characteristics of both styles
Post-Processing Techniques
Common post-processing in Praat:
1. ADD REVERB:
Filter (formula): ~ self + 0.3*self[col-100] + 0.15*self[col-200]
2. ADD COMPRESSION:
Formula: ~ if abs(self) > 0.5 then 0.5*sign(self) else self fi
3. ADD SATURATION:
Formula: ~ tanh(self * 1.5)
4. ADD SWING:
For every even sixteenth: shift by 0.01-0.03s
5. ADD FILTER SWEEP:
Formula: ~ self * sin(2*pi*1000*x) (example: 1000Hz modulation)
Troubleshooting Common Issues
Problem: Pattern sounds robotic/metronomic
Cause: Too perfect timing, no humanization
Solutions:
Cause: Too perfect timing, no humanization
Solutions:
- Use lower groove_density for more space
- Generate stereo version for natural variation
- Add slight timing variations manually after generation
- Use more varied source material
Problem: Sounds cut off abruptly
Cause: Clip_max_length too short, or sounds placed too close together
Solutions:
Cause: Clip_max_length too short, or sounds placed too close together
Solutions:
- Increase Clip_max_length (0.15-0.25s)
- Increase Min_silence_between_events
- Use sparser groove_density
- Increase Release_time for smoother fadeouts
Problem: Stereo version sounds phasey/weird
Cause: Left/right channels too different
Solutions:
Cause: Left/right channels too different
Solutions:
- Use mono version instead
- Manually align some hits between channels
- Use similar random seed for both channels (not implemented in current version)
- Apply mild stereo widening instead of full independent generation
Problem: Classification errors (hi-hats as snares, etc.)
Cause: Source material has ambiguous spectral content
Solutions:
Cause: Source material has ambiguous spectral content
Solutions:
- Use cleaner source material
- Pre-filter audio (high-pass for hi-hats, etc.)
- Manually verify and adjust classification thresholds in code
- Use multiple source files with clear examples of each sound type
Creative Applications
Beyond drum patterns:
- Glitch/IDM: Use very short Clip_max_length (0.02-0.05s) for stutter effects
- Polyrhythms: Generate patterns at different tempos and layer
- Granular textures: High density + very short clips = granular synthesis
- Educational tool: Demonstrate different rhythmic styles
- Composition starter: Generate pattern, then compose melody/harmony to fit
- Sound design: Use non-percussive sounds to create "rhythmic textures"