Tab Gener8or: MP3 ➔ Guitar Tabs & MIDI 🎸➡️📄

Got a riff stuck in your head or need to decode a guitar part from an MP3? Tab Gener8or is a tool for that.

Using Spotify Basic Pitch for note detection and Tayuya for tab-generation, Tab Gener8or analyzes recordings and generates corresponding tablature (tabs) and MIDI files. Useful for:

• Capturing Ideas: Nail down riffs before they vanish.
• Learning by Ear: Get a starting point for transcribing.
• Sharing Parts: Pass along guitar ideas in a standard format.
• AI Music Exploration: See how machines interpret audio.

How's it Work? ✨

Tab Gener8or links together two key open-source libraries:

1. Basic Pitch (from Spotify): Analyzes the audio, detects notes and timing, and outputs a MIDI file. It's the digital ear.
2. Tayuya: Takes that MIDI data and translates it into playable guitar tablature, estimating string and fret positions. It's the digital fingers.

Getting Started Locally

Want to run it yourself? Here's the setup:

1. Clone the Repo: Get the code onto your local machine.

   git clone https://github.com/mgd1984/tab-gener8or.git
   cd tab-gener8or

2. Set Up a Virtual Environment: Keep dependencies tidy. Create and activate a virtual environment.

   # Create (use python3 if needed)
   python -m venv venv
   # Activate
   # macOS/Linux: 
   source venv/bin/activate
   # Windows:
   # venv\Scripts\activate

   (Your prompt should now start with (venv))

3. Install Dependencies: Inside the active (venv), install the required libraries:

   pip install -r requirements.txt

   (This might take a moment, especially downloading TensorFlow for Basic Pitch).

4. Run the App: With the (venv) still active, launch the Streamlit interface:

   streamlit run app.py

   This should open the app in your default web browser (usually http://localhost:8501).

5. Generate Tabs:

   • Upload an MP3 with guitar audio.
   • Hit "Convert to Guitar Tab".
   • Let the AI process.
   • Check out the generated tab and download the .txt or .mid file.

Tips for Better Results & Tinkering 🛠️

• Input Quality Matters: Cleaner audio (less noise/effects, isolated guitar) generally yields better note detection.
• Single Instrument: Basic Pitch works best analyzing one instrument at a time. Feed it isolated guitar tracks if possible.
• Complexity Limits: Very fast, distorted, or complex polyphonic passages can challenge the AI. Clearer tones and simpler lines often transcribe more accurately.
• Command Line Option: If you prefer, use main.py directly (ensure venv is active):

  python main.py path/to/your/audio.mp3

  (Outputs .mid and .txt files in the same directory).
• Code Dive:
  • app.py: The Streamlit UI, audio processing (process_audio), tab formatting (format_tab_for_display).
  • main.py: The script for command-line use.
  • requirements.txt: The list of Python dependencies.

Project Structure

tab-gener8or/
├── app.py                  # Main Streamlit application UI and logic
├── main.py                 # Command-line interface script
├── requirements.txt        # Project dependencies
├── runtime.txt             # Python version target for deployment
├── Dockerfile              # For building a Docker container
├── Procfile                # For Heroku deployment
├── .slugignore             # Files to ignore for Heroku slugs
├── .gitignore              # Files for Git to ignore
├── assets/                 # Static assets (images, etc.)
│   └── images/
└── venv/                   # Virtual environment directory (if created)

Deployment Notes ☁️

• Streamlit Cloud: Convenient for sharing. Push to GitHub, connect the repo. Heads-up: As of this writing, Streamlit Cloud's default Python (3.12) conflicts with basic-pitch==0.4.0's dependencies. Deployment requires ensuring the environment uses Python 3.9 or 3.10.
• Docker: Dockerfile provided for containerized deployment if you need more environment control.

Contributing

Ideas and improvements welcome via Pull Requests.

License

MIT License.

Acknowledgements 🙏

Built using these great open-source projects:

• Basic Pitch by Spotify
• Tayuya by @vipul-sharma20
• Streamlit