Harmonic-Frontier-Audio/Kalimba_Preview

Harmonic Frontier Audio – Kalimba (Preview, v0.9)

A high-fidelity world instrument dataset designed for AI training, music research, and creative sound development.

Kalimba (Preview), created by Harmonic Frontier Audio, provides a compact reference set demonstrating the quality, formatting, and metadata conventions used in the World Resonance Gallery in the Harmonic Frontier Audio catalogue.

---

🔎 Summary

This dataset provides high-quality, rights-cleared recordings of the Kalimba — a lamellophone producing sound through the plucking of tuned metal tines mounted on a resonant body.
The recordings highlight the instrument’s clear attack transients, inharmonic overtone structure, and resonant decay characteristics, making them valuable for AI music modeling, percussive timbre analysis, idiophonic synthesis, and cross-cultural sound design research.

Developed by Harmonic Frontier Audio, this preview follows The Proteus Standard™ for dataset provenance, transparency, and ethical AI use.
Learn more about the Proteus Standard → https://harmonicfrontieraudio.com/proteus-standard

Full dataset details and licensing information are available at: https://harmonicfrontieraudio.com/datasets/kalimba

It offers clean, consistent audio files capturing the Kalimba across single-note articulations, interval patterns, chordal gestures, and controlled plucking techniques.

If you find this dataset useful, please consider giving it a 🤍 on Hugging Face to help others discover it.

---

🎶 About the Kalimba

The Kalimba is a lamellaphone from the Mbira family of instruments, originating in sub-Saharan Africa.
It consists of metal tines attached to a resonating wooden body, played by plucking the tines with the thumbs.
The sound is at once percussive and melodic — metallic transients followed by rich resonant decay.

Kalimbas are often tuned to diatonic or pentatonic scales, though alternate tunings exist across African regions.
This preview dataset was recorded on a 17-key C major Kalimba, offering single-note, dyadic, and arpeggiated gestures across the full playable range.
The goal was to capture the instrument’s core articulations neutrally — not tied to any specific cultural or stylistic context — providing universal training material for AI and creative sound development.

---

📂 Contents

Audio Files (.wav)

• Recorded at 96 kHz / 24-bit WAV format
• Exported as mono
• Fade-ins and fade-outs of 3–5 ms for transient consistency
• DC offset minimized
• No EQ, compression, or post-processing applied
• High-pass filtered at ~40 Hz to remove subsonic rumble

Categories in this Preview

1. Sustain
   • Single tines plucked and allowed to decay naturally
2. Dyad
   • Two tines plucked simultaneously to form harmonic dyads
3. Arpeggio
   • Sequential plucks outlining triadic shapes
4. Chord
   • Three-note simultaneous strikes forming simple triads
5. Rhythm
   • Repeated tine plucks at moderate tempo, emphasizing percussive character

---

Metadata (.csv)
Includes structured fields for file name, category, content, note pitch, Kalimba tab number (where applicable), microphone, channel configuration, sample rate, bit depth, recording chain, and dataset version.

---

🎤 Recording Notes

• Recorded in a treated studio environment using a dual-mic setup blended to mono:
  • Front Mic (transients): Oktava MK-012 (small-diaphragm condenser), 8″ above the tines
  • Rear Mic (resonance): RØDE NT1-A (large-diaphragm condenser), 6″ behind the sound holes
  • Recording chain: Oktava MK-012 + RØDE NT1-A → Zoom F8n Pro
• Captured at 96 kHz / 32-bit float, rendered as 96 kHz / 24-bit mono WAV for release.
• Transient integrity preserved; minor ambient noise retained for natural realism.

---

🌈 Spectrogram Preview

Below is a spectrogram showing the Kalimba’s characteristic harmonic overtones and resonant decay pattern:

---

🎧 Demonstration Audio

This repository includes three audio examples for listening and comparison:

1. Dataset in musical context (mixed & mastered)
A brief musical excerpt demonstrating the Kalimba Preview dataset integrated into a musical arrangement, highlighting expressive timbre and real-world usability.

Track: “Kalimba Example”
Producer: Blake Pullen
Source Audio: Harmonic Frontier Audio – Kalimba (Preview)

2. Raw dataset recording
An unprocessed audio excerpt taken directly from the dataset’s audio files, representing the actual licensed source material without musical context or processing.

3. Fully AI-generated reference (no dataset audio)
A comparison track generated entirely by an AI music model without using any Harmonic Frontier Audio recordings, included solely to illustrate current generative limitations.

---

📜 Demo & Licensing Note

The Demos/ folder contains demonstration material only.
Only the raw recordings in the dataset audio files constitute the licensed dataset.
AI-generated reference tracks are not part of the dataset and are not licensed for training or reuse.

---

🪶 Production Notes

• Instrumental accompaniment in the contextual demo was generated using Suno (Pro plan) under a commercial license.
• All dataset recordings were performed, recorded, and produced by Blake Pullen for Harmonic Frontier Audio.

---

⚡ Usage

This preview dataset is designed for:

• Evaluation of Harmonic Frontier Audio dataset structure and fidelity
• Testing AI and DSP systems that model or classify acoustic instruments
• Creative sound design, instrument synthesis, and timbre modeling research

👉 Note: This is not a full dataset.
The complete Harmonic Frontier Audio dataset for Kalimba will include:

• Full dynamic range and velocity layers
• Expanded tuning systems and gesture types
• Additional playing styles (rolls, damping, harmonics)

---

💡 Full Dataset Availability

This is a preview pack of the Kalimba Dataset.
The complete dataset — with extended articulations and dynamic layers — will be available for licensing.

For licensing inquiries:
📩 info@harmonicfrontieraudio.com

---

📥 How to Use This Dataset in Python

You can load the Parquet-converted version of this dataset directly with the datasets library:

from datasets import load_dataset

dataset = load_dataset(
    "Harmonic-Frontier-Audio/Kalimba_Preview",
    split="train"
)

print(dataset)

⚙️ Note: Parquet conversion and load_dataset() support will be available within 2–3 days of publication.

---

🔗 Explore More from Harmonic Frontier Audio

• Scottish Smallpipes (Preview)
• Highland Bagpipes (Preview)
• Irish Tin Whistle in D (Preview)
• Subharmonic Phonation / Vocal Fry (Preview)
• Kalimba (Preview)
• Kazoo (Preview)
• Overtone Singing (Preview)

(All datasets follow The Proteus Standard™ for ethical dataset provenance and licensing.)

---

📜 License

Released under CC BY-NC 4.0.

• Free for non-commercial use, testing, and research.
• Commercial licensing available via Harmonic Frontier Audio.
• A formal rights declaration is included in this dataset bundle.

---

📧 Contact

Harmonic Frontier Audio
📩 info@harmonicfrontieraudio.com
🌐 https://harmonicfrontieraudio.com/

---

🔮 Future Roadmap

This preview release is part of the Harmonic Frontier Audio – World Resonance Gallery.
Upcoming planned datasets include:

• Kalimba (Full Dataset)
• Didgeridoo
• Shakuhachi

Over time, Harmonic Frontier Audio will expand the World Resonance Gallery alongside its Extended Vocal Techniques Series and Celtic Constellations Series — creating the first unified library of ethical, rights-cleared world and human sound datasets for AI training, synthesis, and sound design.

---

🗒️ Release Notes

Version 0.9 (Nov. 2025) – Initial Preview Pack release for Kalimba.
See CHANGELOG.md for detailed version history.

---

Citation

If you use this dataset in your research, please cite:

Pullen, B. (2025). Kalimba Dataset (Preview) [Data set]. Harmonic Frontier Audio. Zenodo.
https://doi.org/10.5281/zenodo.17588097

ORCID: https://orcid.org/0009-0003-4527-0178

BibTeX

@dataset{pullen_2025_kalimba_preview,
  author       = {Blake Pullen},
  title        = {Kalimba Dataset (Preview)},
  year         = {2025},
  publisher    = {Harmonic Frontier Audio},
  version      = {0.9},
  doi          = {10.5281/zenodo.17588097},
  url          = {https://doi.org/10.5281/zenodo.17588097}
}