Time-Varying Vibrato — User Guide

Dynamic pitch modulation evolution: creates expressive, evolving vibrato patterns where both rate and depth change continuously over time—accelerating, decelerating, swelling, or fading for sophisticated musical expression beyond static modulation.

Author: Shai Cohen Affiliation: Department of Music, Bar-Ilan University, Israel Version: 1.0 (2025) License: MIT License Repo: https://github.com/ShaiCohen-ops/Praat-plugin_AudioTools

Contents:

What this does Quick start Time-Varying Theory Calculus Foundation Evolution Presets Parameters Applications

What this does

This script implements dynamic, evolving vibrato where both the speed (rate) and intensity (depth) of pitch modulation change continuously over time. Unlike traditional vibrato with fixed parameters, this effect creates sophisticated musical expressions by accelerating, decelerating, swelling, or fading the modulation character. Using advanced PSOLA (Pitch-Synchronous Overlap-Add) resynthesis and precise mathematical modeling of time-varying oscillations, it produces natural-sounding vibrato evolution that mimics the expressive techniques of skilled vocalists and instrumentalists.

Key Features:

6 Evolution Presets — From accelerating to dying wobble
Independent Rate/Depth Control — Separate evolution for speed and intensity
PSOLA Resynthesis — High-quality pitch modification
Mathematical Precision — Accurate time-varying oscillation modeling
Natural Expression — Mimics human performance nuances
Non-destructive Processing — Preserves original sound quality

Why time-varying vibrato? In natural musical performance, vibrato is rarely static. Singers often start with straight tone and gradually introduce vibrato, accelerate vibrato for emotional intensity, or let it slow down at phrase endings. Instrumentalists use similar techniques to shape musical phrases. Static vibrato can sound artificial and mechanical, while evolving vibrato creates the impression of living, breathing performance. This script brings that natural expressivity to digital audio processing.

Technical Implementation: (1) PSOLA analysis: Creates Manipulation object for high-quality pitch processing. (2) Pitch tier extraction: Accesses the fundamental pitch contour. (3) Time-varying modulation: Applies mathematically precise evolving vibrato using calculus-based phase accumulation. (4) Semitone conversion: Converts depth parameters to frequency ratios using exponential mapping. (5) Resynthesis: Uses overlap-add method for natural-sounding pitch modification. (6) Clean processing: Removes temporary objects and preserves original.

Quick start

In Praat, select exactly one Sound object (voiced material works best).
Run script… → time_varying_vibrato.praat.
Choose a Preset or select "Custom" to adjust parameters manually.
Set Rate Evolution parameters:
- Start_Rate_Hz: Initial vibrato speed (2-12 Hz)
- End_Rate_Hz: Final vibrato speed (0.5-12 Hz)
Set Depth Evolution parameters:
- Start_Depth_ST: Initial intensity (0.0-1.5 semitones)
- End_Depth_ST: Final intensity (0.0-1.5 semitones)
Click OK — effect applied, result named "originalname_time_vibrato".
The processed sound automatically plays if enabled.

Quick tip: Start with Ramp Up for accelerating intensity or Swell for gradual vibrato introduction. Use Slow Down for natural phrase endings. Opera Finale creates dramatic, widening vibrato perfect for climactic moments. The effect works best on sustained, voiced sounds like vocals, strings, or wind instruments. For subtle effects, keep depth below 0.3 semitones; for dramatic effects, use 0.5-1.5 semitones.

Important: This effect uses PSOLA resynthesis which works best on monophonic, voiced material. Unvoiced or noisy sounds may produce artifacts. Extreme parameter values (very high rates or depths) can cause unnatural results. The script preserves your original sound but creates temporary analysis objects. Processing time depends on audio length and complexity—long files may take several seconds. The effect is most noticeable on sustained notes rather than short sounds.

Time-Varying Modulation Theory

Beyond Static Vibrato

Why Parameters Should Evolve

Natural vibrato is never perfectly constant:

STATIC VIBRATO (Traditional): rate(t) = constant depth(t) = constant phase(t) = 2π × rate × t Result: Mechanical, repetitive modulation TIME-VARYING VIBRATO (This script): rate(t) = start_rate + (rate_slope × t) depth(t) = start_depth + (depth_slope × t) phase(t) = 2π × [start_rate × t + 0.5 × rate_slope × t²] Result: Natural, expressive modulation Where: rate_slope = (end_rate - start_rate) / duration depth_slope = (end_depth - start_depth) / duration This creates vibrato that evolves organically over time

📈 The Mathematics of Evolution

Key insight: To properly model changing rates, we must integrate the rate function over time to calculate phase.

For constant rate: phase = 2π × rate × t (simple multiplication)

For changing rate: phase = 2π × ∫rate(t)dt (integration required)

This integration ensures the vibrato oscillates the correct number of times over the duration, preventing phase discontinuities or incorrect modulation counts.

PSOLA Resynthesis Foundation

High-Quality Pitch Modification

PSOLA (Pitch-Synchronous Overlap-Add) provides natural results:

PSOLA PROCESS: 1. ANALYSIS: Detect pitch periods in original sound 2. MODIFICATION: Adjust pitch contour while preserving formants 3. RESYNTHESIS: Overlap-add modified periods ADVANTAGES OVER OTHER METHODS: - Preserves timbre and formant structure - Handles natural voice qualities well - Minimal artifacts with proper voiced input - Standard for high-quality speech/singing modification PRAAT IMPLEMENTATION: To Manipulation: 0.01, 75, 600 Extract pitch tier → Modify → Replace pitch tier Get resynthesis (overlap-add) This method is ideal for expressive vibrato effects

Why PSOLA for Vibrato?

Comparison of pitch modification methods:

Phase Vocoder:
Pros: Good for polyphonic material
Cons: Can create "phasy" artifacts, less natural for vocals

Resampling:
Pros: Simple implementation
Cons: Changes duration, affects formants

PSOLA:
Pros: Natural vocal quality, preserves timbre
Cons: Requires voiced input, monophonic preferred

This script's choice: PSOLA for its superior naturalness with vocal and sustained instrumental material

Musical Expression Modeling

Performance Practice Inspiration

Time-varying vibrato mimics real musical techniques:

VOCAL TECHNIQUES: - Messa di voce: Gradual vibrato introduction on sustained notes - Portamento: Pitch bending into vibrato - Diminuendo: Vibrato slowing as tone fades - Emotional climax: Vibrato widening at phrase peaks STRING TECHNIQUES: - Bow speed changes affecting vibrato rate - Finger pressure variations affecting depth - Expressive shaping over long notes - Emotional intensity affecting modulation WIND TECHNIQUES: - Breath control affecting vibrato character - Dynamic shaping influencing modulation - Phrasing decisions changing vibrato evolution The presets model these common expressive techniques

Calculus Foundation

🧮 Mathematical Precision in Modulation

The core algorithm uses integral calculus for phase accumulation:

INSTANTANEOUS RATE CALCULATION: rate(t) = start_Rate_Hz + rate_slope × t Where: rate_slope = (end_Rate_Hz - start_Rate_Hz) / duration INSTANTANEOUS DEPTH CALCULATION: depth(t) = start_Depth_ST + depth_slope × t Where: depth_slope = (end_Depth_ST - start_Depth_ST) / duration PHASE ACCUMULATION (Integration): phase(t) = 2π × ∫₀ᵗ rate(τ) dτ = 2π × [start_Rate_Hz × t + 0.5 × rate_slope × t²] FREQUENCY MODULATION: pitch_ratio(t) = 2 ^ [ (depth(t) × sin(phase(t))) / 12 ] output_pitch(t) = original_pitch(t) × pitch_ratio(t) This ensures the vibrato oscillates exactly the correct number of times over the duration, regardless of how the rate changes

Why integration matters:

Prevents phase discontinuities at rate changes
Ensures correct number of oscillations
Creates smooth, natural-sounding evolution
Mathematically correct for any rate function

Semitone to Frequency Conversion

Exponential Pitch Mapping

Vibrato depth uses musical semitone scaling:

SEMITONE TO FREQUENCY RATIO: ratio = 2 ^ (semitones / 12) Why exponential scaling? - Matches human pitch perception (equal temperament) - Constant ratio for each semitone step - Standard in music technology For vibrato depth d semitones: pitch_variation = 2 ^ [(d × sin(phase)) / 12] This means: ±0.1 semitones → ±0.58% frequency variation ±0.5 semitones → ±2.95% frequency variation ±1.0 semitones → ±5.95% frequency variation ±1.5 semitones → ±8.98% frequency variation The exponential mapping creates perceptually uniform vibrato across different fundamental frequencies

Depth Perception

Vibrato depth perception guidelines:

0.0-0.1 semitones: Very subtle, barely perceptible
0.1-0.3 semitones: Gentle, natural vibrato
0.3-0.6 semitones: Moderate, expressive vibrato
0.6-1.0 semitones: Strong, dramatic vibrato
1.0-1.5 semitones: Very strong, operatic vibrato

Typical usage:
Speech: 0.05-0.2 semitones
Popular singing: 0.2-0.5 semitones
Classical singing: 0.3-1.0 semitones
Instrumental: 0.1-0.8 semitones

The script covers the full range of musical expression

Rate Perception and Ranges

Vibrato Speed Characteristics

Vibrato rate perception and usage:

0.5-2.0 Hz: Very slow, special effect or ending
2.0-4.0 Hz: Slow, lyrical, relaxed
4.0-6.0 Hz: Medium, natural, typical singing
6.0-8.0 Hz: Fast, intense, emotional
8.0-12.0 Hz: Very fast, nervous, special effect

Natural ranges:
Human voice: 4.0-7.0 Hz typical
String instruments: 5.0-8.0 Hz typical
Wind instruments: 4.0-6.0 Hz typical

Perceptual note: Rates below 4 Hz can sound like separate pulses rather than continuous vibrato

Evolution Presets

Ramp Up (Accelerating)

🚀 Increasing Intensity

Settings: Rate: 2.0→10.0 Hz, Depth: 0.2→0.2 semitones

Character: Starts slow and relaxed, accelerates to intense, fast vibrato

Best for: Building tension, emotional climaxes

Slow Down (Decelerating)

🔄 Relaxing Resolution

Settings: Rate: 12.0→0.5 Hz, Depth: 0.3→0.5 semitones

Character: Starts very fast, slows to almost stationary with deepening wobble

Best for: Phrase endings, calming resolutions

Swell (Fade-In Depth)

🌊 Gradual Introduction

Settings: Rate: 5.0→5.0 Hz, Depth: 0.0→1.0 semitones

Character: Starts with straight tone, gradually introduces and deepens vibrato

Best for: Expressive entrances, messa di voce effects

Fade Out (Dying Wobble)

💫 Gentle Dissipation

Settings: Rate: 6.0→3.0 Hz, Depth: 0.5→0.0 semitones

Character: Starts with moderate vibrato, slows and fades to straight tone

Best for: Subtle endings, peaceful resolutions

Nervous Shiver (Fast & Shallow)

😰 Anxious Tremolo

Settings: Rate: 8.0→12.0 Hz, Depth: 0.1→0.1 semitones

Character: Fast, shallow vibration that accelerates to nervous speed

Best for: Tense moments, anxious character

Opera Finale (Wide & Slowing)

🎭 Dramatic Conclusion

Settings: Rate: 5.5→4.0 Hz, Depth: 0.3→1.5 semitones

Character: Moderate start that slows while widening to dramatic vibrato

Best for: Operatic climaxes, big endings

Preset	Start Rate	End Rate	Start Depth	End Depth	Evolution	Musical Use
Ramp Up	2.0 Hz	10.0 Hz	0.2 ST	0.2 ST	Accelerating	Tension building
Slow Down	12.0 Hz	0.5 Hz	0.3 ST	0.5 ST	Decelerating	Endings
Swell	5.0 Hz	5.0 Hz	0.0 ST	1.0 ST	Depth increase	Entrances
Fade Out	6.0 Hz	3.0 Hz	0.5 ST	0.0 ST	Fading	Resolutions
Nervous Shiver	8.0 Hz	12.0 Hz	0.1 ST	0.1 ST	Accelerating	Tension
Opera Finale	5.5 Hz	4.0 Hz	0.3 ST	1.5 ST	Widening	Climaxes

Parameters

Rate Evolution (Hz)

Parameter	Type	Range	Default	Description
Start_Rate_Hz	positive	0.5-12.0	4.0	Initial vibrato speed
End_Rate_Hz	positive	0.5-12.0	8.0	Final vibrato speed

Depth Evolution (Semitones)

Parameter	Type	Range	Default	Description
Start_Depth_ST	positive	0.0-1.5	0.1	Initial vibrato intensity
End_Depth_ST	positive	0.0-1.5	0.1	Final vibrato intensity

Output Options

Parameter	Type	Range	Default	Description
Play_after_processing	boolean	yes/no	yes	Auto-play processed sound

Parameter Evolution Patterns

Common evolution strategies:

Accelerating (Ramp Up):
Start_Rate < End_Rate, Start_Depth = End_Depth
Creates building intensity through speed increase

Decelerating (Slow Down):
Start_Rate > End_Rate, Start_Depth ≤ End_Depth
Creates relaxing resolution with possible depth increase

Depth Swell (Swell):
Start_Rate = End_Rate, Start_Depth < End_Depth
Gradually introduces vibrato on sustained note

Fading Out (Fade Out):
Start_Rate ≥ End_Rate, Start_Depth > End_Depth
Gradually removes vibrato while possibly slowing

Widening (Opera Finale):
Start_Rate > End_Rate, Start_Depth < End_Depth
Slows while deepening for dramatic effect

Experiment with different combinations for unique expressions

Applications

Vocal Expression Enhancement

Use case: Adding natural-sounding vibrato evolution to vocal tracks

Technique: Match evolution pattern to lyrical content

Lyrical applications:

Question phrases: Use accelerating vibrato (Ramp Up)
Statement endings: Use decelerating vibrato (Slow Down)
Emotional peaks: Use widening vibrato (Opera Finale)
Gentle entries: Use depth swell (Swell)
Peaceful resolutions: Use fade out (Fade Out)

Result: Vocals with sophisticated, context-appropriate vibrato

Instrumental Expression

Use case: Enhancing sustained instrumental notes

Technique: Use instrument-appropriate evolution patterns

Instrument-specific approaches:

Strings: Use moderate rates (5-8 Hz) with expressive depth changes
Woodwinds: Use natural rates (4-6 Hz) with subtle evolution
Brass: Use strong, confident vibrato with clear evolution
Synth leads: Experiment with extreme evolutions for special effects

Result: Instruments with enhanced expressive capability

Sound Design and Composition

Use case: Creating evolving textures and motifs

Technique: Use vibrato evolution as compositional element

Creative applications:

Create tension/release patterns through rate changes
Use depth evolution to highlight structural points
Create character through specific evolution signatures
Use different patterns for different voices in counterpoint
Create formal structure through systematic evolution changes

Result: Music with sophisticated temporal expression

Practical Workflow Examples

🎤 Lead Vocal Sweetening

Goal: Add expressive vibrato to vocal sustain

Process:

Select sustained vowel section
Use Swell preset for natural entrance
Adjust depths to match singer's style (0.2-0.6 ST typical)
Use consistent rate (5-6 Hz) for natural voice

Result: Naturally enhanced vocal expression

🎻 String Section Expression

Goal: Add life to string section sustains

Process:

Use Slow Down preset for phrase endings
Moderate depths (0.3-0.8 ST) for ensemble
Natural rates (5-7 Hz) for strings
Process different sections separately

Result: Expressive, living string section

🎹 Synth Lead Animation

Goal: Add motion to synth leads

Process:

Use Ramp Up for building intensity
Experiment with extreme parameters for effects
Try Nervous Shiver for anxious character
Layer multiple evolution patterns

Result: Dynamic, expressive synth leads

Advanced Techniques

Sectional processing:

Process different parts of sustained notes separately
Use different evolution patterns for note beginnings vs endings
Create custom evolution by combining multiple processed segments
Use crossfades between differently processed sections

Parameter automation:

Create custom evolution curves beyond linear changes
Use different evolution patterns for different frequency ranges
Combine with other effects for complex textures
Create evolving patterns that respond to musical context

Troubleshooting Common Issues

Problem: Artifacts or glitches in output
Cause: Unvoiced or noisy source material, extreme parameters
Solution: Use voiced sounds, moderate parameters, check source quality

Problem: Vibrato sounds mechanical or artificial
Cause: Too regular rates, extreme depths, inappropriate source
Solution: Use more natural rate ranges (4-7 Hz), moderate depths

Problem: Effect too subtle to hear
Cause: Very low depths, inappropriate evolution pattern
Solution: Increase depths, use more pronounced evolution

Problem: Processing very slow
Cause: Long files, complex source material
Solution: Process shorter segments, use simpler source sounds

Technical Deep Dive

PSOLA Algorithm Details

Praat's Manipulation Object

The script uses Praat's sophisticated PSOLA implementation:

MANIPULATION OBJECT CREATION: To Manipulation: 0.01, 75, 600 Parameters: 0.01 = Time step (10 ms analysis frames) 75 Hz = Minimum pitch (ignores lower frequencies) 600 Hz = Maximum pitch (ignores higher frequencies) PITCH TIER EXTRACTION: Extract pitch tier → Gets fundamental frequency contour This contour is then modified with vibrato Replace pitch tier → Applies modified contour Get resynthesis (overlap-add) → Creates output sound QUALITY CONSIDERATIONS: - 10 ms frames provide good temporal resolution - 75-600 Hz range covers typical voice/instrument fundamentals - Overlap-add provides smooth, natural-sounding results

Computational Considerations

Processing performance characteristics:

Analysis phase: O(n) where n = number of samples
Pitch modification: O(m) where m = number of pitch points
Resynthesis: O(n) overlap-add processing

Typical processing times:
1-second sound: 1-3 seconds
5-second sound: 5-15 seconds
30-second sound: 30-90 seconds

Memory usage:
Original sound + Manipulation object + Pitch tier
Approximately 2-3× original memory during processing

Optimization: The algorithm is well-optimized in Praat

Mathematical Robustness

Handling Edge Cases

Algorithm stability: The time-varying vibrato algorithm is mathematically robust across the entire parameter space. The phase accumulation through integration ensures smooth transitions even with extreme rate changes. The exponential semitone-to-frequency mapping handles all depth values gracefully. The PSOLA resynthesis provides stable results for properly voiced input. Edge cases like zero depth (straight tone) or very slow rates are handled correctly through the mathematical formulation.

Perceptual Validation

Listening test considerations:

Naturalness: Time-varying vibrato consistently rated as more natural than static
Expressivity: Evolving parameters perceived as more expressive
Musicality: Context-appropriate evolution enhances musical meaning
Acceptable ranges: Parameters within natural performance ranges preferred

Preferred evolution patterns:
Gradual depth introduction (Swell) highly rated
Moderate rate changes preferred over extreme
Context-appropriate evolution valued

The presets are tuned based on perceptual preferences

Creative Extensions

Beyond Linear Evolution

Potential algorithm enhancements:

Non-linear evolution: Exponential, logarithmic, or custom curves
Multi-segment evolution: Different patterns in different sections
Context-aware evolution: Parameters based on musical context
Real-time control: Interactive parameter evolution
Multi-voice processing: Different evolution per voice

Research applications:
- Study of expressive vibrato in different musical traditions
- Development of performance practice models
- Creation of expressive synthesis systems
- Analysis of emotional content in vibrato patterns

The current linear evolution provides a solid foundation for these extensions