"تری خودی میں اگر انقلاب ہو پیدا"
— کلیاتِ اقبال: اردو صوتیاتی ڈیٹا سیٹ برائے مصنوعی ذہانت

🎙️ Kulliat-e-Iqbal-TTS Dataset (v1.0) 🇵🇰

Kulliat-e-Iqbal-TTS is a high-fidelity Urdu speech dataset specifically curated for training and fine-tuning Text-to-Speech (TTS) models. Part of the Shaheen AI Project, it comprises 9,336 high-quality audio samples paired with precise transcriptions from the masterworks of Allama Muhammad Iqbal.

This dataset is designed to enable modern neural speech synthesis models to recite classical Urdu poetry with authentic rhythmic structures and linguistic nuances.

---

🌟 Key Highlights

• Literary Breadth: Includes verses from Armaghan-e-Hijaz, Bal-e-Jibril, Zarb-e-Kaleem, and the Iqbal TTS corpus.
• TTS Optimized: All audio normalized to 22050Hz, Mono, 16-bit PCM for out-of-the-box compatibility with VITS, Piper, Glow-TTS, and Coqui.
• Linguistic Precision: Cleaned transcriptions utilizing the specialized vocabulary of Iqbaliyat to ensure high-quality prosody.
• Large Scale: Approximately 11.8 hours of validated speech data, making it a premier open-source resource for Urdu poetic TTS.

---

📊 Dataset Composition

The dataset is a consolidated and cleaned merge of four major literary sources:

Category	Source	Samples	Focus
Armaghan-e-Hijaz	Armaghan Auto-Whisper	516	Persian & Urdu Quatrains
Bal-e-Jibril	Bal-Jibril Corrected	2,183	High-energy Ghazals
Iqbal TTS	General Collection	4,987	Rhythmic Verse Recitation
Zarb-e-Kaleem	Zarb Corrected	1,650	Philosophical Declarative Prose

---

🛠️ Technical Specifications

• Total Rows: 9,336
• Format: file_name,text (Standard CSV)
• Audio Codec: WAV (PCM)
• Sampling Rate: 22,050 Hz
• Bit Depth: 16-bit
• Channels: Mono (Single Channel)
• Language Code: ur (Urdu)

---

💎 Features for Researchers

Feature	Detail
Total Duration	~11.8 Hours
Average Sample Length	4.2 Seconds
Metadata Delimiter	Comma (,) with Headers
Script Support	Full Nastaliq / UTF-8 Encoding

---

🚀 Quick Start (Hugging Face Datasets)

To use this dataset in your Python environment:

from datasets import load_dataset

# Load the dataset from Khurram123
dataset = load_dataset("Khurram123/Kulliat-e-Iqbal-TTS")

# Preview the first sample
print(dataset['train'][0])

# To listen to audio in a Jupyter Notebook:
import IPython.display as ipd
ipd.Audio(dataset['train'][0]['audio']['array'], rate=22050)