FireRedVAD-ASR-SRT 🎙️🎬

Integrated FireRedVAD with multiple ASR models, supporting single-file or batch SRT subtitle generation.

中文 | English | 日本語

---

🌟 Core Features

• 🎯 Multi-model Support: SenseVoiceSmall, Paraformer, Fun-ASR-Nano, Fun-ASR-MLT-Nano, Granite-Speech, GLM-ASR-Nano
• 🎭 Multi-language Interface: English, Chinese, Japanese (easily extensible)
• 📝 Batch Processing: Single-file or batch transcription
• ⚡ High Performance: Optimized for both CPU and GPU acceleration
• 🎛️ Flexible Configuration: Adjustable VAD parameters and model settings
• 📊 Rich Output: SRT subtitle format with timestamps

🚀 Quick Start

1. Environment Setup

Use uv to create a virtual environment (recommended):

uv venv --python 3.12

2. Install Dependencies

Install using uv:

uv pip install -r requirements.txt

3. Model Configuration

Download and configure models:

• SenseVoiceSmall: Automatically downloads
• FunAsr-nano: Automatically downloads
• GLM-Asr-nano: Automatically downloads
• Granite-Speech: Supported languages: English, French, German, Spanish, Portuguese and Japanese. Download from modelscope or Huggingface
• VAD Model: FireRedVAD folder

Important: All models can be taken from the cache folder and placed into the project folder.

4. Hardware Acceleration

🎮 NVIDIA GPU (CUDA):

uv pip install torch torchaudio --index-url https://download.pytorch.org/whl/cu126 （Please install the appropriate version based on the CUDA version installed on your system.）

💻 CPU Only:

uv pip install torch torchaudio --index-url https://download.pytorch.org/whl/cpu

5. Run the Application

uv run main.py

📁 Project Structure

FireRed/
├── 📄 main.py                 # Main program
├── 📄 ctc.py                  # CTC module
├── 📄 model.py                # Fun-ASR model code
├── 📄 requirements.txt        # Python dependencies
├── 📁 asr/                    # ASR package sources
├── 📁 FireRedVAD/             # Voice activity detection model files
├── 📁 FunAudioLLM/            # Fun-ASR local model folders
├── 📁 ibm-granite/            # Granite Speech model files
├── 📁 iic/                    # Additional speech model folders
├── 📁 locales/                # Language translations
├── 📁 tools/                  # Utility scripts and tools
├── 📁 utils/                  # Translation utilities
├── 📁 ZhipuAI/                # Additional local model folders
├── 📄 README.md               # This file
├── 📄 README_ZH.md            # Chinese version
└── 📄 README_JA.md            # Japanese version

🎯 Usage Tips

Important: When performing batch transcription, always try single-file transcription first to find the optimal VAD parameters and ensure accurate sentence segmentation. Please prioritize adjusting the VAD parameter Min Silence Frames to get better sentence segmentation.

🤝 Contributing

We welcome all kinds of contributions! Feel free to:

• 🐛 Report bugs
• 💡 Suggest features
• 🌍 Add translations
• 🔧 Improve code

🙏 Acknowledgements

This project relies on the following open-source projects:

• Fun-ASR — ASR model pipeline, Apache 2.0

See NOTICE.md for full license texts and copyright notices.

📄 License

This project is open-sourced under the Apache License 2.0 - see the LICENSE file for details.

📸 Interface Preview