U-CLONE is an AI-based Urdu Voice Cloning System developed as a Final Year Project (FYP). The system focuses on cloning a person’s voice in the Urdu language using Deep Learning and Text-to-Speech (TTS) technologies.
It demonstrates how artificial intelligence can synthesize natural-sounding Urdu speech in the voice of a target speaker.
The goal of this project was to build a personalized voice synthesis system capable of generating Urdu speech that mimics a given speaker’s voice. This was achieved through speech data preprocessing, acoustic model training, and waveform generation using state-of-the-art deep learning models.
- 🗣️ Voice Cloning: Clone any speaker’s voice using provided audio samples.
- 📝 Text-to-Speech Conversion: Convert Urdu text into natural-sounding speech.
- ⚙️ Voice Customization: Adjust pitch, tone, and speed.
- 💻 Web-Based Interface: Simple and interactive system design.
-
Dataset Preparation
- Collected Urdu speech samples from multiple speakers.
- Preprocessed data for training (noise removal, normalization, segmentation).
-
Model Development
- Used Tacotron 2 for speech synthesis.
- Integrated Griffin-Lim for realistic audio generation.
- Fine-tuned for Urdu phonetic structure and script processing.
-
Evaluation
- Measured Mean Opinion Score (MOS) for speech naturalness.
- Validated speaker similarity and intelligibility.
-
Deployment
- Designed a user interface for Urdu text input and voice generation.
| Category | Tools / Frameworks |
|---|---|
| Programming | Python |
| Deep Learning | PyTorch, TensorFlow |
| Speech Synthesis | Tacotron 2, Griffin-Lim |
| NLP Processing | Urdu Text Normalization, Tokenization |
| Interface | HTML, CSS, Flask |
| Others | NumPy, Pandas, Matplotlib |
This repository contains the supporting documentation of the project:
- 📘 Final Report (Thesis Document)
- 📊 Presentation Slides (PPT)
- 📄 Project Abstract and Overview Files
⚠️ Note: Source code and model training files are not included in this repository. This repository is intended for academic presentation and documentation only.
Project Title: U-CLONE: Urdu Voice Cloning System Type: Final Year Project (FYP) Status: Completed (March 2025)
Team Members:
- 🧑💻 Areeba Shehbaz
- 👩💻 Surooj Virk
Supervisor:
- 🎓 Dr. Farah Adeeba
Department: BS Software Engineering Institution: University of Engineering and Technology Lahore
- Successfully cloned Urdu voices with high naturalness.
- Demonstrated model’s ability to generalize across speakers.
- Developed a working prototype for text-to-speech generation in Urdu.