Skip to content

README_DEV.md

Latest commit

 

History

History
89 lines (62 loc) · 2.49 KB

README_DEV.md

File metadata and controls

89 lines (62 loc) · 2.49 KB

VoiceTransor

VoiceTransor is an open-source speech-to-text and text assistant.
It provides a simple workflow to transcribe audio locally using Whisper, process text with AI, and export results in multiple formats.

✨ Features

  • Import audio files
  • Local transcription with Whisper (supports resume from interruption)
  • AI-powered text processing
  • Export results as TXT / PDF
  • Cross-platform: Windows, macOS

🚀 Installation

Prerequisites

  • Python 3.10+
  • FFmpeg installed and available in PATH
  • Virtual environment (recommended)

Configure AI Text Processing

VoiceTransor uses Ollama for local AI-powered text processing. This keeps your data completely private without sending it to the cloud.

Quick Setup:

  1. Windows: Run scripts\setup\install_ollama.bat from project root (automatic installation)
  2. Manual installation: Download from ollama.com/download
  3. Start service: Run ollama serve in a terminal
  4. Download a model: Run ollama pull llama3.1:8b

For detailed setup instructions, see OLLAMA_SETUP_GUIDE.md (中文版).

Recommended Models:

  • llama3.1:8b - English (default, ~4.7GB)
  • qwen2.5:7b - Balanced Chinese/English (~4.4GB)
  • gemma2:9b - High quality (~5.4GB)

System Requirements:

  • GPU mode: NVIDIA GPU with 8GB+ VRAM (recommended)
  • CPU mode: 16GB+ RAM (slower but works)

Setup

git clone https://github.com/leonshen/VoiceTransor.git
cd VoiceTransor
pip install -r requirements.txt

Windows GPU (CUDA) setup

If you want Whisper to use an NVIDIA GPU on Windows:

  1. Uninstall any existing CPU-only PyTorch wheels inside your virtualenv: 
    pip uninstall torch torchvision torchaudio -y
  2. Install the matching CUDA wheels (examples below use CUDA 12.1; pick the build that matches your driver): 
    pip install torch==2.3.0+cu121 torchvision==0.18.0+cu121 torchaudio==2.3.0+cu121 \
    --index-url https://download.pytorch.org/whl/cu121
  3. Verify CUDA is detected before launching VoiceTransor: 
    python -c "import torch; print(torch.cuda.is_available(), torch.version.cuda)"
    The command should print True and a CUDA version string.

Run

Make sure the virtual environment is activated

python -m app.main

📧 Contact

For support or collaboration: voicetransor@gmail.com

📜 License

MIT License. See LICENSE for details.