🎙️ Nojoin

⚖️ Legal Disclaimer

Important: You are responsible for complying with all applicable laws in your jurisdiction regarding the recording of conversations. Many jurisdictions require the consent of all parties before a conversation can be recorded. By using Nojoin, you acknowledge that you will use this software in a lawful manner. The developers of Nojoin assume no liability for any unlawful use of this application.

✨ Features

• System Audio Recording: Simultaneously record what you hear (system output) and what you say (microphone input).
• Local-First Transcription: Uses OpenAI's Whisper to generate accurate transcripts of your recordings, running entirely on your machine.
• Offline Speaker Diarization: Automatically identifies who spoke when using Pyannote for offline speaker diarization. No cloud connection required for core processing.
• Comprehensive Speaker Management: Relabel, merge, and manage speakers per recording. Build a global speaker library for consistent naming across all your meetings.
• LLM-Powered Insights:
  • Generate concise, actionable meeting notes and summaries.
  • Ask questions about your meeting transcript in a chat-style Q&A interface.
  • Supports OpenAI, Google Gemini, and Anthropic models (requires your own API key).
• Powerful Search and Organization:
  • Full-text search across all meeting transcripts.
  • High-performance find and replace text within a single transcript or across all of them.
  • Organize recordings with custom tags.
• Data Management & Backup:
  • Complete backup and restore system creating portable zip files with all your data.
  • Database-first architecture for faster operations and reliable data storage.
  • Non-destructive restore that merges with existing data.
• Automatic Updates: Built-in update checking and installation system with user-configurable preferences.
• Full Control: Manage recordings, view transcripts, and configure settings like transcription models and audio devices through a modern UI.

Manual Setup

1. Prerequisites:

   • Python 3.11.9: This project requires Python 3.11.9 SPECIFICALLY due to PyTorch compatibility issues. You can download it directly from the Python website. The typical caveats of running multiple Python versions/installations apply.

     Alternatively, you can install it using your system's package manager:

     • Windows (winget):

       winget install Python.Python.3.11

     • macOS (Homebrew):

       brew install [email protected]

     • Linux (Debian/Ubuntu):

       sudo apt update && sudo apt install python3.11 python3.11-venv

   • ffmpeg: ffmpeg is required for audio processing and must be available in your system's PATH.

     • Windows (winget):

       winget install ffmpeg

     • macOS (Homebrew):

       brew install ffmpeg

     • Linux (Debian/Ubuntu):

       sudo apt update && sudo apt install ffmpeg

   • Optional (for GPU Acceleration): NVIDIA GPU with the CUDA 12.8 Toolkit installed. See the "GPU Acceleration" section for more details.

2. Clone the repository:

   git clone https://github.com/Valtora/Nojoin

   Switch to the newly cloned directoy:

   cd Nojoin

3. Create and activate a virtual environment: It is highly recommended to use a virtual environment to manage project dependencies. Create the virtual environment:

   python -m venv .venv

   Activate the environment:

   • Windows (PowerShell/CMD):

     .venv\Scripts\Activate

   • macOS/Linux (bash/zsh):

     source .venv/bin/activate

4. Install Dependencies

   The dependencies are listed in requirements.txt. However, the PyTorch dependency is highly system-specific (CPU/GPU, OS, architecture). It is crucial to install the correct PyTorch version for your machine before installing the other packages.

   • Step 1: Install PyTorch Visit the PyTorch website to determine the correct installation command for your system configuration. For example:

     • Windows/Linux (CPU-only):

       pip install torch torchvision torchaudio

     • Windows/Linux (CUDA 12.8):

       pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu128

     • macOS (ARM/M-series chip):

       pip install torch torchvision torchaudio

   • Step 2: Install Remaining Dependencies Once PyTorch is installed, you can install the rest of the required packages. The requirements.txt file is configured for a CUDA-enabled setup, but pip will use your manually installed PyTorch version and skip reinstalling it.

     pip install -r requirements.txt

5. Run Nojoin: Make sure your virtual environment is active before running the application:

   python Nojoin.py

GPU Acceleration (CUDA Support)

Nojoin supports GPU acceleration for transcription and diarization using CUDA. To enable GPU support:

• You must have an NVIDIA GPU and the CUDA Toolkit version 12.8 installed.
• Compatible NVIDIA drivers must be installed.
• PyTorch must be installed with CUDA 12.8 support (see the installation steps above).
• CUDA availability is automatically detected by Nojoin. If available, you can select "cuda" as the processing device in the Settings dialog.
• If CUDA is not detected, only CPU processing will be available.

Note: CUDA 12.8 is the only supported version for now. Other versions may not work correctly. Saturn Cloud wrote a helpful troubleshooting guide if you have CUDA issues.

☕ Buy Me a Coffee

If Nojoin is useful to you please consider buying me a coffee as a way to support the project.