lahja-sa-huba-v1

**Lahja SA Huba 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 Huba 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. - 🇸🇦 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 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-Huba-v2`** pip install transformers[torch] from transformers import VitsModel, AutoTokenizer import torch model = VitsModel.from_pretrained("wasmdashai/lahja-sa-huba-v1") tokenizer = AutoTokenizer.from_pretrained("wasmdashai/lahja-sa-huba-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) Lahja SA Huba 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 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 📧 <modelasg@gmail.com> • 🌍 [Website](https://anas.lahjai.net) • 🤗 [Hugging Face](https://huggingface.co/wasmdashai) • 💻 [GitHub](https://github.com/asgmodel/vits-pro) • 🔗 [LinkedIn](https://www.linkedin.com/in/anas-al-tawil/) 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. 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. @misc{lahja_sa_Huba_v1, title={Lahja SA Huba V1: Saudi Arabic Text-to-Speech Model}, author={Anas Al-Tawil}, year={2026}, publisher={Hugging Face}, url={https://huggingface.co/wasmdashai/lahja-sa-Huba-v1} } **Developed with ❤️ by WasmAI** Building the Future of Arabic Artificial Intelligence © 2026 WasmAI. All rights reserved.