Skip to content

About

Best-shot audio-to-sheet-music with full MIDI editing, playback, and printable scores. No AI.

Topics

Resources

Stars

1 star

Watchers

0 watching

Forks

Latest commit

 

History

2 Commits

Folders and files

Repository files navigation

ScoreForge

Download ScoreForge Download ScoreForge

Best-shot audio-to-sheet-music with a full MIDI editor, playback, and printable scores. No AI anywhere in the pipeline — pure classical DSP.

Ships as .AppImage (Linux), .app / .dmg (macOS), and .exe (Windows).

What it does

  • Load an audio file (mp3 / wav / flac / ogg / m4a / aac / opus / webm).
  • Vocal reduce (karaoke-style L−R subtract) or vocal isolate (center emphasis).
  • Transcribe a dominant melody line to MIDI notes using one of several pitch detectors (YIN, AMDF, Dynamic Wavelet, Autocorrelation).
  • Edit everything in a canvas piano roll:
    • draw / erase / move / resize notes
    • add or remove tracks and swap instrument per track
    • change tempo, time signature, snap, key, zoom
    • keyboard: arrows nudge pitch/time, Shift+Up/Down jumps octaves, Ctrl/Cmd+A selects all, Delete removes
  • Play the project back through a built-in Tone.js synth (one instrument preset per track).
  • Render sheet music on screen with VexFlow — clef, key signature, time signature all configurable.
  • Export MIDI, MusicXML, or print the sheet music to PDF.

Honest limits

ScoreForge does its best shot. Here is what "best shot" means in practice.

  • Monophonic bias. The transcription assumes one dominant melody line in each frame. Dense polyphony (full band mixes) will not come out "semi-perfect" — the detector will drift to whatever is loudest or clearest from frame to frame. The editor is what turns that first pass into a usable score.
  • Vocal reduction is center-channel subtraction. It works when the vocal is panned center on a plain stereo mix; it fails on wide-stereo modern mixes and will also eat anything else sitting in the center (kick, snare, bass).
  • No AI stem separation, by design.

For a cleaner first pass, start with a solo instrument or a vocal line, or run Vocal reduce / Vocal only before Transcribe.

Run from source

npm install
npm start

Build native installers

npm run build:linux     # AppImage
npm run build:mac       # dmg
npm run build:win       # NSIS installer (.exe)
npm run build:all       # all three

Output lands in dist/.

Keyboard shortcuts

key action
Arrow Up / Down nudge selected notes by 1 semitone
Shift + Arrow Up / Down nudge by 1 octave
Arrow Left / Right nudge by one snap step
Delete / Backspace delete selected notes
Ctrl/Cmd + A select all notes in active track
Ctrl/Cmd + wheel zoom
Shift + wheel horizontal scroll
wheel vertical scroll

Project model

A project is a plain object held in memory:

{
  tempo: 120,
  beatsPerBar: 4,
  unit: 4,
  tracks: [
    {
      id: 1,
      name: 'Melody',
      instrument: 'Piano',
      color: '#4ea1ff',
      muted: false,
      notes: [
        { pitch: 72, startSec: 0.0, durationSec: 0.5, velocity: 96 },
        ...
      ]
    }
  ]
}

MIDI export uses @tonejs/midi. MusicXML export is hand-rolled and loads in MuseScore, Finale, Sibelius, Dorico, Flat.io and most notation software.

Legal

Transcribing copyrighted recordings for private study is one thing; distributing scores derived from them is another. This app is a tool — what you do with its output is on you.

License

MIT.

About

Best-shot audio-to-sheet-music with full MIDI editing, playback, and printable scores. No AI.

Topics

Resources

Stars

1 star

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages