🎵 AI Music Generator — CodeAlpha Internship

A professional AI-powered music generation desktop application built with LSTM Deep Learning

CodeAlpha AI Internship — Task 3

---

📌 About The Project

AI Music Generator is a fully GUI-based desktop application that uses Long Short-Term Memory (LSTM) neural networks to learn music patterns from the MAESTRO dataset and generate original music compositions. The app allows users to train the AI model, generate music with customizable settings, and play it back — all from a beautiful, modern interface.

This project was built as part of the CodeAlpha AI Internship Program to demonstrate real-world application of deep learning in creative AI.

---

🎬 Demo

Train the model → Generate music → Play it back instantly!

The app features:

• 🎓 Train Tab — Train LSTM model with live loss graph
• 🎼 Generate Tab — Generate music with custom settings
• ▶️ Playback Tab — Play generated music with waveform visualizer

---

✨ Features

🤖 AI & Machine Learning

• LSTM-based deep learning model built with PyTorch
• Trained on MAESTRO Dataset (1,276 professional piano MIDI files)
• Extracts and learns note sequences and chord patterns
• Generates completely original music compositions

🎨 Professional GUI

• Modern dark/light theme with CustomTkinter
• Animated splash screen on launch
• Dark/Light mode toggle
• Scrollable, responsive layout

🎵 Music Generation Controls

• Notes to Generate — slider (50 to 500 notes)
• Creativity (Temperature) — controls randomness (0.1 to 2.0)
• BPM (Tempo) — slider (60 to 200 BPM)
• Instrument Selector — Piano, Violin, Guitar, Flute, Cello, Choir and more
• Genre Style — Classical, Jazz, Ambient, Pop, Cinematic
• Key / Scale — C Major, A Minor, D Minor and more

📊 Training Dashboard

• Live loss graph updates during training
• Training progress bar with epoch counter
• Detailed training logs

▶️ Playback & Export

• Animated waveform visualizer while playing
• Play / Stop controls
• Copy file path to clipboard
• Open output folder directly
• Delete generated files

---

🛠️ Tech Stack

Technology	Purpose
Python 3.14	Core programming language
PyTorch	LSTM model training & inference
music21	MIDI parsing and note extraction
pretty_midi	MIDI file generation
CustomTkinter	Modern GUI framework
pygame-ce	MIDI playback
NumPy	Numerical computations
Pillow	Image processing

---

📁 Project Structure

CodeAlpha_MusicGenerator/
│
├── 📁 gui/
│   ├── main_window.py      # Main app window + splash screen
│   ├── train_tab.py        # Training UI + loss graph
│   ├── generate_tab.py     # Music generation UI
│   └── playback_tab.py     # Playback + waveform visualizer
│
├── 📁 model/
│   ├── train.py            # LSTM model + training logic
│   ├── generate.py         # Music generation logic
│   ├── music_model.pth     # Saved trained model (after training)
│   └── notes.pkl           # Extracted notes cache
│
├── 📁 data/
│   ├── 📁 2004/            # MAESTRO MIDI files by year
│   ├── 📁 2006/
│   ├── 📁 2008/
│   └── ...
│
├── 📁 output/              # Generated MIDI files saved here
├── App.py                  # Entry point — run this!
└── requirements.txt        # All dependencies

---

🚀 Getting Started

Prerequisites

• Python 3.10, 3.11, or 3.14
• pip package manager
• Windows / Mac / Linux

Installation

1. Clone the repository

git clone https://github.com/YourUsername/CodeAlpha_MusicGenerator.git
cd CodeAlpha_MusicGenerator

2. Create virtual environment

python -m venv venv
venv\Scripts\activate        # Windows
source venv/bin/activate     # Mac/Linux

3. Install dependencies

pip install torch music21 pretty_midi pygame-ce customtkinter pillow numpy

4. Download MAESTRO Dataset

• Go to: https://magenta.tensorflow.org/datasets/maestro
• Download MIDI only version (~57MB)
• Extract and place year folders (2004, 2006...) inside data/

5. Run the app

python App.py

---

🎮 How To Use

Step 1 — Train The Model

1. Open the app → click Train Model tab
2. Set epochs (20 recommended) and MIDI files (10 recommended)
3. Click 🚀 Start Training
4. Watch the live loss graph update
5. Wait for "Training Complete!" message

⚠️ Training is a one-time process. The model saves automatically.

Step 2 — Generate Music

1. Click Generate Music tab
2. Adjust your settings:
   • Choose instrument, genre, key, BPM
   • Set creativity level
   • Enter output filename
3. Click 🎼 Generate Music
4. Wait a few seconds for generation

Step 3 — Play & Export

1. Click Playback tab
2. Click 🔄 Refresh to load your file
3. Select your generated file
4. Click ▶️ Play and enjoy the waveform visualizer!

---

🎵 Recommended Settings For Best Results

Style	BPM	Creativity	Instrument	Key
Classical	72	0.7	Acoustic Grand Piano	A Minor
Drama OST	65	0.7	Violin	D Minor
Jazz	110	1.3	Vibraphone	D Minor
Ambient	60	0.9	Choir Aahs	C Major
Cinematic	85	1.0	Violin	E Minor
Pop	128	1.2	Electric Piano	G Major

---

📸 Screenshots

---

🤖 How The AI Works

MIDI Files → Note Extraction → Sequence Preparation
     ↓
LSTM Neural Network Training
     ↓
Pattern Learning (notes, chords, rhythms)
     ↓
Music Generation (new sequences)
     ↓
MIDI File Output

The LSTM model learns the statistical patterns in music — which notes typically follow which other notes, chord progressions, and musical phrases. During generation, it uses these learned patterns to create new, original compositions.

---

👨‍💻 Developer

Ubaidullah Waheed

• 🎓 CodeAlpha AI Internship
• 💼 LinkedIn: [https://www.linkedin.com/in/ubaidullah-waheed-a119ba383/]
• 🐙 GitHub: [https://github.com/UbaidullahWaheed]

---

🏢 About CodeAlpha

CodeAlpha is a leading software development company dedicated to driving innovation and excellence across emerging technologies. This project was built as part of their AI internship program.

🌐 Website: www.codealpha.tech

---

📄 License

This project is licensed under the MIT License — feel free to use and modify it!

---

⭐ If you found this project helpful, please give it a star! ⭐

Built with ❤️ by Ubaidullah Waheed — CodeAlpha AI Internship 2026