Update README.md

README.md

@@ -8,50 +8,31 @@ tags:
 - text-generation-inference
 - TTS
 ---
-# 🎙️ MamreVoice: State-of-the-Art Hebrew TTS
 
-<p align="center">
-  <img src="https://raw.githubusercontent.com/maxmelichov/mamre-site/main/static/images/logo.png" alt="Mamre Logo" width="200"/>
-</p>
+# 🎙️ MamreVoice: State-of-the-Art Hebrew TTS
 
-[**Project Site**](https://maxmelichov.github.io/mamre-site/) | [**GitHub**](https://github.com/maxmelichov) | [**Demo**](https://huggingface.co/spaces/maxmelichov/mamre-demo)
+[**Project Site**](https://maxmelichov.github.io/mamre-site/) | [**GitHub**](https://github.com/maxmelichov)
 
-MamreVoice is a high-performance, end-to-end Text-to-Speech (TTS) engine specifically optimized for the Hebrew language. By combining the **DiffMamba** architecture with advanced Hebrew diacritization, it achieves studio-quality synthesis with remarkably low latency.
+MamreVoice is a high-performance, end-to-end Text-to-Speech (TTS) and Voice Conversion system specifically optimized for the Hebrew language. Built on the innovative **DiffMamba** architecture, it delivers expressive, clear, and lifelike synthetic speech.
 
 ## ✨ Features
 
-* **DiffMamba Architecture**: Utilizes a 2B parameter model merging Diffusion generation with Mamba (SSM) efficiency.
-* **Zero-Shot Voice Cloning**: Clone any Hebrew voice with just a 5-10 second reference clip.
-* **Advanced Diacritization**: Built-in support for **Phonikud**, ensuring correct pronunciation of Niqqud (vowels).
-* **Production Ready**: Optimized for ONNX and TensorRT, making it viable for real-time applications.
-* **Multi-Lingual Foundation**: While specialized for Hebrew, the model maintains high-fidelity across various acoustic environments.
-
----
-
-## 🚀 Performance Benchmarks
-
-MamreVoice is optimized for extreme throughput without sacrificing quality.
-
-| Hardware | RTF (Real-Time Factor) | Description |
-| :--- | :--- | :--- |
-| **NVIDIA RTX 4090** | ~1,500x | Near-instant generation |
-| **NVIDIA RTX 3090** | 1,260x | High-end workstation |
-| **CPU (i7-12700K)** | 35x | Faster than real-time |
-| **Apple M2 Max** | 25x | Efficient mobile performance |
+* **Expressive Hebrew Synthesis**: Specialized optimization for Hebrew prosody, rhythm, and naturalness.
+* **Voice Conversion**: High-quality voice cloning and identity preservation.
+* **Phonetics-Aware**: Seamless integration with **Phonikud** for accurate Hebrew diacritization (Niqqud) and vowelization.
+* **Efficiency**: Designed for low computational overhead, enabling real-time performance on modern Nvidia GPU hardware.
 
 ---
 
 ## 🛠️ Technical Overview
 
-### The Pipeline
-The model processes text through a specialized Hebrew pipeline:
-1.  **Normalization**: Handling numbers, dates, and symbols.
-2.  **Phonikud**: Automatic vowelization and phonetic transcription.
-3.  **DiffMamba Backbone**: The core model predicts the mel-spectrogram using a State Space Model (SSM) backbone.
-4.  **Vocoder**: High-fidelity reconstruction into 24kHz or 44.1kHz audio.
+### Architecture
+MamreVoice utilizes **DiffMamba**, which addresses the computational heavy-lifting of traditional diffusion models by using Mamba blocks. This allows for long-context handling and faster inference without sacrificing audio quality.
 
+### Components
+1.  **Text Processing**: Advanced normalization and diacritization via [Phonikud](https://github.com/thewh1teagle/phonikud).
+2.  **BackBone Model**: DiffMamba-based Backbone
+---
 
 
-### Requirements
 ```bash
-pip install torch torchaudio phonikud onnxruntime-gpu