wasmdashai / lahja-sa-ahmad-v1

Lahja SA Ahmad V1

Lahja SA Ahmad V1 is a production-ready Arabic Text-to-Speech (TTS) model developed by WasmAI and optimized for the Saudi dialect. Built on the VITS architecture, the model generates natural, expressive, and human-like speech while preserving Saudi pronunciation, rhythm, and linguistic characteristics.

Designed for both research and enterprise applications, Lahja SA Ahmad V1 enables developers to integrate realistic Arabic speech synthesis into conversational AI, voice assistants, accessibility technologies, educational platforms, robotics, customer service systems, and other intelligent applications.

---

✨ Features

• 🇸🇦 Natural Saudi Arabic speech synthesis
• 🎙️ Human-like pronunciation and expressive intonation
• ⚡ Fast inference with low latency
• 🧠 End-to-end VITS architecture
• 🤖 Production-ready deployment
• 🤗 Fully compatible with Hugging Face Transformers

---

🏗️ Architecture

The model is based on Variational Inference with Adversarial Learning for End-to-End Text-to-Speech (VITS) .

Its architecture combines:

• Transformer Text Encoder
• Variational Autoencoder (VAE)
• Flow-based Prior Network
• Stochastic Duration Predictor
• HiFi-GAN Neural Decoder

Base model:

wasmdashai/vits-ar-sa-Ahmad-v2

---

🚀 Installation

pip install transformers[torch]

---

💻 Usage

from transformers import VitsModel, AutoTokenizer
import torch

model = VitsModel.from_pretrained("wasmdashai/lahja-sa-ahmad-v1")
tokenizer = AutoTokenizer.from_pretrained("wasmdashai/lahja-sa-ahmad-v1")

text = "السلام عليكم ورحمة الله وبركاته"

inputs = tokenizer(text, return_tensors="pt")

with torch.no_grad():
    output = model(**inputs)

waveform = output.waveform.cpu().numpy().reshape(-1)

from IPython.display import Audio

Audio(waveform, rate=model.config.sampling_rate)

---

🎯 Applications

Lahja SA Ahmad V1 is suitable for:

• Conversational AI
• Voice Assistants
• Interactive Chatbots
• Customer Service Automation
• Educational Platforms
• Accessibility Solutions
• Robotics
• Smart Devices
• IVR Systems
• Enterprise AI Applications

---

📚 Dataset

The model was fine-tuned using the LAHJA-API-Data dataset, specifically curated for Saudi Arabic speech synthesis to improve pronunciation, fluency, and dialect consistency.

Dataset:

https://huggingface.co/datasets/wasmdashai/LAHJA-API-Data

---

🌐 Contact

📧 [email protected] • 🌍 Website • 🤗 Hugging Face • 💻 GitHub • 🔗 LinkedIn

---

🏢 About WasmAI

WasmAI develops enterprise-grade Artificial Intelligence solutions focused on:

• 🤖 Generative AI
• 🧠 Large Language Models (LLMs)
• 🗣️ Arabic Speech AI
• 👁️ Computer Vision
• 🤝 AI Agents
• 🔍 Retrieval-Augmented Generation (RAG)
• ☁️ Enterprise AI Platforms

Our mission is to build scalable, production-ready AI technologies that accelerate digital transformation across industries.

---

🙏 Acknowledgements

This project builds upon the excellent work of the open-source community, including VITS , HiFi-GAN , Transformers , PyTorch , tts-arabic , Finetune HF VITS , and Bert-VITS2 . We sincerely thank all contributors for advancing speech synthesis research.

---

📖 Citation

@misc{lahja_sa_ahmad_v1,
  title={Lahja SA Ahmad V1: Saudi Arabic Text-to-Speech Model},
  author={Anas Al-Tawil},
  year={2026},
  publisher={Hugging Face},
  url={https://huggingface.co/wasmdashai/lahja-sa-ahmad-v1}
}

---

````