AI-Music-Workbench

AI-Music-Workbench is a collaborative repository for generating, transforming, and experimenting with algorithmic music using AI systems such as Codex and Claude. The project focuses on text-based music representations including ABC notation, MIDI, and symbolic music formats.

Live Site

• https://aleph-beth.github.io/AI-Music-Workbench/

PowerShell MIDI Scripts

1) Dynamic Classical String Quartet

• Script: generate_quartet_midi.ps1
• Output: output/quatuor_classique_plus_4min_dynamique.mid
• Features: string quartet (about 4 minutes), progressive dynamics, phrase accents.

Command:

powershell -ExecutionPolicy Bypass -File .\generate_quartet_midi.ps1

2) Electro / Techno Remix (Long Notes)

• Script: generate_electro_remix_midi.ps1
• Output: output/remix_electro_techno_sons_longs.mid
• Features: 128 BPM techno groove, electronic kick/hat, synth bass, sustained pads, long lead.

Command:

powershell -ExecutionPolicy Bypass -File .\generate_electro_remix_midi.ps1

Project Structure

• generate_quartet_midi.ps1: generates a format-1 MIDI classical string quartet.
• generate_electro_remix_midi.ps1: generates a format-1 MIDI electro/techno remix track.
• output/: generated MIDI files and static manifest.

Web MIDI Interface (GitHub Pages)

This repository includes a static web interface compatible with GitHub Pages:

• index.html: UI layout (MIDI selection, loading, instrument mapping, playback controls)
• style.css: responsive styling
• app.js: frontend logic (manifest loading, MIDI parsing, instruments, playback scheduling)
• output/midi-index.json: MIDI manifest used for file listing

Architecture Summary

1. MIDI manifest loading:

• The frontend fetches output/midi-index.json.
• No runtime directory listing is used.
• .mid and .midi entries populate the "Choose a MIDI file" dropdown.

2. MIDI parsing:

• app.js uses @tonejs/midi (browser ESM via CDN).
• Tracks are displayed with:
  • track number
  • optional track name
  • note count, channel, duration, and range

3. Per-track instrument mapping:

• A predefined instrument catalog is centralized in INSTRUMENT_DEFINITIONS.
• Each track has an independent instrument selector.
• A TrackInstrumentPlayer abstraction maps each track to a playable instrument object.

4. Synchronized playback:

• Tone.js Transport drives browser playback.
• Notes are scheduled per track with Tone.Part in sync.
• Controls: Load MIDI, Play, Stop.

Updating the MIDI Manifest

When adding or removing MIDI files in output/, update output/midi-index.json:

{
  "files": [
    "example_1.mid",
    "example_2.midi"
  ]
}

GitHub Pages Deployment

1. Commit and push to github.com/aleph-beth/AI-Music-Workbench.
2. Open GitHub Settings > Pages.
3. Select Deploy from a branch.
4. Choose branch main (or another branch) and folder / (root).
5. Save.

Expected URL:

• https://aleph-beth.github.io/AI-Music-Workbench/

All frontend paths are relative (for example output/...), so the site works correctly under the repository subpath on GitHub Pages.