VIP2VIP: A Multi-Dimensional Pipeline for Holistic Prosodic Visualization

VIP2VIP (Visualize Intonation and Prosody for Voices of Italian Poets) is an open-source Python-based pipeline designed for the integrated visualization of fundamental frequency ($f_0$), intensity, and syllabic rate. While traditional speech research often privileges $f_0$ alone, this tool renders a multi-dimensional display to reveal how energy and tempo interact with pitch to cue prominence and phrasing.

How It Works

The pipeline compresses three acoustic dimensions into a single coordinated display:

• $f_0$ Contour: Renders a smoothed pitch trace using either LOWESS or GAM (Generalized Additive Models).
• Intensity: Encoded as line thickness (instantaneous intensity); thicker segments indicate higher energy.
• Syllabic Rate (SR): Encoded as a perceptually ordered color ramp (e.g., green for slower rates, red for faster rates) projected directly onto the $f_0$ trace.

Key Features

• Superpositional Framework: Designed to visualize global phrase-level trends alongside local accentual movements.
• Auto Flex Smoothing: Automatically adjusts smoothing parameters based on the number of syllables in a phrase to maintain visual comparability.
• Micro-prosodic Robustness: Sampling is restricted to voiced phonetic intervals to limit artifacts from voiceless consonants.
• Reproducible Outputs: Parameter choices are logged for each unit, and data can be exported as CSV for statistical analysis or PNG for publications.

---

Requirements & Annotations

The pipeline is designed to run on lightweight, reproducible annotations.

Software

• Python environment (Jupyter Notebook).
• Core Libraries: Parselmouth (Praat backend), Matplotlib, statsmodels, and praatio.

Required Annotation Tiers (TextGrid)

Your Praat TextGrids must include the following tiers:

• CP (Chunk/Prosodic Unit): Defines the working window for plotting (e.g., interpausal units).
• syll (Syllables): Phonetic syllable intervals used to derive local and global syllabic rates.
• phon (Voiced Segments): Marks voiced intervals (vowels/voiced sonorants) for precise acoustic sampling.
• Optional: INT for intonational labels and VS for textual verse lines.

---

Quick Start Guide

1. Environment Setup: Ensure you have the required Python libraries installed: pip install parselmouth-praat matplotlib statsmodels praatio
2. Data Preparation: Place your audio files and corresponding TextGrids (with CP, syll, and phon tiers) in your working directory.
3. Run the Pipeline: Open the Jupyter Notebook and execute the visualization function.
4. Configure Smoothing: Choose between LOWESS (default for local curvature) or GAM (for stable global trends) and enable Auto Flex for automatic smoothness scaling.
5. Interactive Selection: Use the UI to batch process files or interactively select specific prosodic units (CPs) for detailed inspection.
6. Export: Save the generated plots as images and extract the acoustic indices to a CSV file for further study.

---

Citation

If you use this pipeline in your research, please cite the original paper:

Lo Iacono, F., & Romano, A. (2026). A Multi-Dimensional Pipeline for Holistic Prosodic Visualization: Integrating f0, Intensity, and Syllabic Rate using VIP2VIP. Speech Prosody 2026, Philadelphia, University of Pennsylvania.

---

License: Open source. For issues or contributions, please visit the official repository.