Project Ideas

Appearance

Algorithmic Music Generator ​

DifficultyAdvanced
Team Size2-3 people
Time~30-40 hours
Demo-ready byStep 5
PrerequisitesJavaScript, basic music theory, audio concepts
Built byAIVA, Amper Music, MuseNet, Sonic Pi

Skills you'll earn: Web Audio API, music theory algorithms, Markov chains, MIDI generation, audio synthesis

Start with a beep. End with a system that composes music from rules.

(Assumes basic JS knowledge. Includes significant time learning Web Audio API, music theory, and DSP concepts.)

Step 1: Play a sound in the browser (~1-2 hours) ​

You want to hear a note. Just one note.

  • One HTML page with a "Play" button
  • Use the Web Audio API: create an AudioContext and an OscillatorNode
  • Set the frequency to 440 Hz (A4)
  • Start and stop the oscillator on button click
  • Add a gain node to control volume

You now have: A browser synthesizer playing a single note.

Step 2: Play a sequence of notes (~2-3 hours) ​

One note isn't music. You need melody.

  • Define a scale as an array of frequencies (C major: C4, D4, E4, F4, G4, A4, B4)
  • Play each note for a fixed duration (200ms), one after another
  • Use setTimeout or schedule with AudioContext.currentTime for precise timing
  • Add a tempo control (BPM)

You now have: A note sequencer.

Step 3: Generate melodies algorithmically (~4-5 hours) ​

Playing a fixed scale isn't composition. You want the computer to create melodies.

  • Implement random note selection from a scale
  • Add rules: prefer stepwise motion (move to adjacent notes), occasional leaps
  • Use a Markov chain: the next note depends on the current note (build a transition probability table)
  • Generate 16-bar melodies

You now have: Algorithmic melody generation.

Step 4: Rhythm generation (~3-4 hours) ​

Every note has the same duration. Real music has rhythm.

  • Define note durations: whole, half, quarter, eighth, sixteenth
  • Assign random durations weighted toward common patterns
  • Add rests (silence) as a possible duration
  • Ensure measures add up to the correct beat count (e.g., 4 beats per bar)

You now have: Rhythmic patterns.

Step 5: Chords and harmony (~4-5 hours) ​

A single melodic line sounds thin.

  • Generate chord progressions using common patterns (I-IV-V-I, ii-V-I)
  • Play chords as multiple simultaneous oscillators
  • The melody should follow the chord tones (or resolve to them on strong beats)
  • Add a bass line that follows the chord roots

You now have: Harmony.

Step 6: Better sound (~4-5 hours) ​

Oscillator beeps sound terrible.

  • Shape the sound with ADSR envelopes (attack, decay, sustain, release) using gain node automation
  • Add different waveforms: sine, square, sawtooth, triangle
  • Add effects: reverb (ConvolverNode), delay, filter (BiquadFilterNode)
  • Layer sounds for richness

You now have: Synthesized instruments.

Step 7: Export (~3-4 hours) ​

  • Record the output using MediaRecorder and the audio context's destination
  • Export as WAV or MP3
  • Generate MIDI output (use a MIDI library) for importing into a DAW

Step 8: Genre presets (~3-4 hours) ​

  • Define rule sets for different styles: ambient, jazz, classical, lo-fi
  • Each genre has its own scales, chord progressions, rhythms, and timbres
  • Let users mix parameters across genres

Step 9: Visual feedback (~3-4 hours) ​

  • Show a piano roll visualization of the generated music
  • Animate notes as they play
  • Display the current chord and scale

Useful Resources ​

Where to go from here ​

  • Machine learning models trained on real music (RNN, transformer)
  • User-guided generation (hum a melody, the system harmonizes it)
  • Multi-instrument arrangements
  • Live performance mode (generate in real time, never repeating)
  • Collaborative composition