Vinu - VieNeu-TTS

Welcome to the Vinu repository! This project contains a modified and enhanced version of VieNeu-TTS, a powerful Vietnamese Text-To-Speech (TTS) and Voice Cloning system.

🚀 Features

• High-Quality Vietnamese TTS: Natural sounding Vietnamese text-to-speech with various preset voices.
• Voice Cloning: Clone custom voices using reference audio.
• SRT Subtitle Processing (New!): Synthesize audio directly from .srt subtitle files. The system automatically reads the text without speaking the timestamps, and inserts precise silences to ensure the generated audio aligns perfectly with your subtitle timings!
• XPU/Intel GPU Support: Optimized execution scripts for Intel Iris/Arc GPUs as well as standard Nvidia GPUs.

🛠️ Getting Started

Prerequisites

Make sure you have uv installed to manage the Python environment seamlessly.

Running the Web UI

1. For Standard/Nvidia GPUs: Run the main Gradio application:

   cd VieNeu-TTS
   uv run python apps/gradio_main.py
   

2. For Intel GPUs (XPU): Simply double-click the run_xpu.bat file located inside the VieNeu-TTS folder, or run:

   cd VieNeu-TTS
   run_xpu.bat
   

📝 How to use the SRT Feature

1. Open the Web UI.
2. Paste your .srt file contents directly into the main text box.
3. Check the box labeled "Xử lý văn bản như file SRT (tự động căn chỉnh độ trễ)" located just below the text input.
4. Select your voice and click Generate! The output will automatically pad silences to match your subtitle timings.

---

Note: Large model weights (.safetensors, .bin) are ignored by git to keep this repository lightweight. The system will automatically download them from HuggingFace upon the first run.